a NATO Advanced Study Institute

LES HOUCHES

SESSION LXXVI 30 July – 31 August 2001 Unity from duality: Gravity, gauge theory and strings L’unité de la physique fondamentale : gravité, théorie de jauge et cordes

Edited by C. BACHAS, A. BILAL, M. DOUGLAS, N. NEKRASOV and F. DAVID

Springer

Les Ulis, Paris, Cambridge Berlin, Heidelberg, New York, Hong Kong, London, Milan, Paris, Tokyo

Published in cooperation with the NATO Scientific Affair Division

ISSN 0924-8099 print edition ISSN 1610-3459 online edition ISBN 3-540-00276-6 ISBN 2-86883-625-9 Springer-Verlag Berlin Heidelberg New York EDP Sciences Les Ulis

This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, re-use of illustrations, recitation, broadcasting, reproduction on microfilms or in other ways, and storage in data banks. Duplication of this publication or parts thereof is only permitted under the provisions of the French and German Copyright laws of March 11, 1957 and September 9, 1965, respectively. Violations fall under the prosecution act of the French and German Copyright Laws. © EDP Sciences; Springer-Verlag 2002 Printed in France

Preface

The 76th session of the Les Houches Summer School in Theoretical Physics was devoted to recent developments in string theory, gauge theories and quantum gravity. As frequently stated, Superstring Theory is the leading candidate for a uniﬁed theory of all fundamental physical forces and elementary particles. This claim, and the wish to reconcile general relativity and quantum mechanics, have provided the main impetus for the development of the theory over the past two decades. More recently the discovery of dualities, and of important new tools such as D-branes, has greatly reinforced this point of view. On the one hand there is now good reason to believe that the underlying theory is unique. On the other hand, we have for the ﬁrst time working (though unrealistic) microscopic models of black hole mechanics. Furthermore, these recent developments have lead to new ideas about compactiﬁcation and the emergence of low-energy physics. While pursuing the goal of uniﬁcation we have also witnessed a dramatic return to the “historic origins” of string theory as a dual model for meson physics. Indeed, the study of stringy black branes has uncovered a surprising relation between string theory and large-N gauge dynamics. This was cristallized in the AdS/CFT correspondence, which has revived the old hope for a string description of the strong interaction. The AdS/CFT correspondence is moreover a prime illustration of the central role of string theory in modern theoretical physics. Much like quantum ﬁeld theory in the past, it provides a fertile springboard for new tools, concepts and insights, which should have ramiﬁcations in wider areas of physics and mathematics. The main lectures of the Les Houches school covered most of the recent developments, in a distilled and pedagogical fashion. Students were expected to have a good knowledge of quantum ﬁeld theory, and of basic string theory at the level, for instance, of the ﬁrst ten chapters of Green, Schwarz and Witten. The emphasis was on acquiring a working knowledge of advanced string theory in its present form, and on critically assessing open problems and future directions. The lectures by Bernard de Wit were a comprehensive introduction to supergravities in diﬀerent dimensions and with various numbers of supersymmetries. Topics covered include the allowed low-energy couplings, duality symmetries, compactiﬁcations and supersymmetry in curved backgrounds.

xxii

Part of this is older material not easily accessible in the literature, and presented here from a modern perspective. Eliezer Rabinovici lectured on supersymmetric gauge theories, reviewing earlier and more recent results for N = 1, 2 and 4 supersymmetries in four dimensions. These results include the structure of the eﬀective lagrangians, non-renormalization theorems, dualities, the celebrated Seiberg-Witten solution and brane engineering of eﬀective gauge theories. M-theory and string dualities were introduced in the lectures by Ashoke Sen. He reviewed the conjectured relations between the ﬁve perturbative string theories, the maximal N = 1 supergravity in eleven dimensions and their compactiﬁcations. He summarized our present-day knowledge of the still elusive fundamental or “M theory”, from which the above theories derive as special limits. More recent topics include non-BPS branes, where duality is of limited (but not zero) use. Philip Candelas gave a pedagogical introduction to the important subject of Calabi Yau compactiﬁcations. He ﬁrst reviewed the older material, and then discussed more recent aspects, including second quantized mirror symmetry, conifold transitions and some intriguing relations to number theory. Unfortunately a written version of his lectures could not be included in this volume. The holographic gauge/string theory correspondence was the subject of the lectures by Juan Maldacena and by Igor Klebanov. Maldacena introduced the conjectured equivalence between string theory in the near-horizon geometries of various black branes and gauge theories in the large Ncolor limit. He focused on the celebrated example of N = 4 four-dimensional super Yang Mills dual to string theory in AdS5 × S5 , and gave a critical review of the existing evidence for this correspondence. He also discussed analogous conjectures in other spacetime dimensions, in particular those relevant to the study of stringy black holes, and of the still elusive little string theory. Igor Klebanov then concentrated on this duality in the phenomenologically more interesting contexts of certain N = 1 and 2 supersymmetric gauge theories in four dimensions. He reviewed the relevant geometries on the supergravity side, which include non-trivial ﬂuxes and fractional branes, and discussed the gravity duals of renormalization group ﬂow, conﬁnement and chiral symmetry breaking. These results have revived and made sharper the old ideas about the “master ﬁeld” of large N gauge theory. The lectures of Michael Green dealt with some ﬁner aspects of string dualities and of the gauge theory/string theory correspondence. He discussed higher derivative couplings in eﬀective supergravity actions, focusing in particular on the contributions of instantons both in string theory and on the

xxiii

Yang Mills side. His lectures also included some introductory material on Dbranes. Unfortunately a written version of his lectures could not be included in this volume. Andrew Strominger gave a detailed introduction to quantum gravity in a de Sitter spacetime. He discussed in particular whether ideas of holography, that have worked well in anti de Sitter, could also be applied in this case. This was one of the more speculative subjects in the school, but a fascinating one not the least because astrophysical observations seem to indicate that we actually live in an accelerating universe. Finally Michael Douglas gave three lectures on D-brane geometry, and in particular on the problem of classifying all N = 1 string-theory vacua, while Alexander Gorsky discussed N = 1 and N = 2 supersymmetric gauge theories and their relation to integrable models. Nikita Nekrasov lectured on open strings and non-commutative gauge theories. Some more advanced and/or topical subjects were covered in the accompanying series of seminars. Seminar speakers included Laurent Baulieu, Mirjam Cvetiˇ, Frank Ferrari, Dan Freedman, Bernard Julia, Peter Mayr, c Soo-Jong Rey, Augusto Sagnotti, Samson Shatashvili, and one of the organizers (C.B.). There was also a lively weekly student seminar and discussion sessions, which contributed greatly to the lively and stimulating atmosphere of the school. Some of the seminar speakers have kindly accepted to contribute to the present volume. In the year that has elapsed since the end of the school there have been further developments in the subject. The pp-waves, which arise as Penrose limits of near-horizon geometries, oﬀer for instance a new line of attack on the important problem of solving string theory in Ramond-Ramond backgrounds. Such developments and others will no doubt make, one day, the present volume obsolete. This is of course no reason for regret – to the contrary we hope that this may happen sooner rather than later, and that the participants of this school will help shape the (non-recognisable?) future form of M theory. Acknowledgements This Les Houches school and the present proceedings were made possible due to the contributions of many individuals and several diﬀerent funding sources. Funding sources included the NATO Scientiﬁc Aﬀairs Division through its Advanced Study Institute program, the European Union through its High-Level Scientiﬁc Conferences program, the Commissariat ` l’Energie a ´ Atomique, the Universit´ Joseph Fourier de Grenoble, and two European e

xxiv

IHP Networks: “Superstring Theory” (HPRN-CT-2000-00122) and “The Quantum Structure of Spacetime” (HPRN-CT-2000-00131). Among the many people contributing to the success of the school, we should mention - the board of the School and in particular Fran¸ois David, who has c worked tirelessly at all diﬀerent stages (funding applications, admissions, running the session, preparation of proceedings) exceeding often the organizers in zeal and energy; - the secretaries Mmes G. D’Henry, I. Leli`vre and B. Rousset (and the e other personnel of the school), who helped solve administrative and everyday problems; and last but not least - the lecturers, for their eﬀorts in presenting hard material in a clear and pedagogical fashion, and also for writing up their lecture notes.

Costas Bachas Adel Bilal Michael Douglas Nikita Nekrasov Fran¸ois David c

CONTENTS

Lecturers Participants Pr´face e Preface Contents

xi xiii xvii xxi xxv

**Lecture 1. Supergravity by B. de Wit
**

1 Introduction 2 Supersymmetry in various dimensions 2.1 The Poincar´ supersymmetry algebra . . . . e 2.2 Massless supermultiplets . . . . . . . . . . . 2.2.1 D = 11 supermultipets . . . . . . . . 2.2.2 D = 10 supermultiplets . . . . . . . 2.2.3 D = 6 supermultiplets . . . . . . . . 2.3 Massive supermultiplets . . . . . . . . . . . 2.4 Central charges and multiplet shortening . . 2.5 On spinors and the R-symmetry group HR

1

3 4 5 7 8 9 13 17 19 22 28 28 32 35 38 43 46 53 55 56 61 66 69

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

3 Supergravity 3.1 Simple supergravity . . . . . . . . . . . . . . . . . . 3.2 Maximal supersymmetry and supergravity . . . . . . 3.3 D = 11 supergravity . . . . . . . . . . . . . . . . . . 3.4 Dimensional reduction and hidden symmetries . . . 3.5 Frames and ﬁeld redeﬁnitions . . . . . . . . . . . . . 3.6 Kaluza–Klein states and BPS-extended supergravity 3.7 Nonmaximal supersymmetry: Q = 16 . . . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

4 Homogeneous spaces and nonlinear sigma models 4.1 Nonlinearly realized symmetries . . . . . . . . . . . . . . 4.2 Geometrical quantities . . . . . . . . . . . . . . . . . . . 4.3 Nonlinear sigma models with homogeneous target space 4.4 Gauged nonlinear sigma models . . . . . . . . . . . . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

xxvi

5 Gauged maximal supergravity in 4 and 5 dimensions 5.1 On E7(7) /SU (8) and E6(6) /U Sp(8) cosets . . . . . . . . 5.2 On ungauged maximal supergravity Lagrangians . . . . 5.3 Electric–magnetic duality and E7(7) . . . . . . . . . . . 5.4 Gauging maximal supergravity; the T -tensor . . . . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

70 72 74 79 84

6 Supersymmetry in anti-de Sitter space 89 6.1 Anti-de Sitter supersymmetry and masslike terms . . . . . . . . . . 93 6.2 Unitary representations of the anti-de Sitter algebra . . . . . . . . 97 6.3 The superalgebras OSp(N |4) . . . . . . . . . . . . . . . . . . . . . 110 7 Superconformal symmetry 116 7.1 The superconformal algebra . . . . . . . . . . . . . . . . . . . . . . 119 7.2 Superconformal gauge theory and supergravity . . . . . . . . . . . 122 7.3 Matter ﬁelds and currents . . . . . . . . . . . . . . . . . . . . . . . 126

**Lecture 2. Supersymmetric Gauge Theories by D.S. Berman and E. Rabinovici
**

1 Introduction 2 Supersymmetric quantum mechanics 2.1 Symmetry and symmetry breaking . . . . . . 2.2 A nonrenormalisation theorem . . . . . . . . 2.3 A two variable realization and ﬂat potentials 2.4 Geometric meaning of the Witten index . . . 2.5 Landau levels and SUSY QM . . . . . . . . . 2.6 Conformal quantum mechanics . . . . . . . . 2.7 Superconformal quantum mechanics . . . . . 3 Review of supersymmetric models 3.1 Kinematics . . . . . . . . . . . . . 3.2 Superspace and chiral ﬁelds . . . . 3.3 K¨hler potentials . . . . . . . . . . a 3.4 F-terms . . . . . . . . . . . . . . . 3.5 Global symmetries . . . . . . . . . 3.6 The eﬀective potential . . . . . . . 3.7 Supersymmetry breaking . . . . . . 3.8 Supersymmetric gauge theories . . 4 Phases of gauge theories

137

141 143 151 152 154 158 159 161 164 165 165 167 169 170 170 172 172 173 179

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

xxvii

5 Supersymmetric gauge theories/super QCD 5.1 The classical moduli space . . . . . . . . . . . 5.2 Quantum moduli spaces . . . . . . . . . . . . 5.3 Quantum moduli space for 0 < NF < NC . . 5.4 Integrating in . . . . . . . . . . . . . . . . . . 5.5 Quantum moduli space for NF ≥ NC . . . . . 5.6 NF = NC . . . . . . . . . . . . . . . . . . . . 5.7 NF = NC + 1 . . . . . . . . . . . . . . . . . . 5.8 Higgs and conﬁnement phases . . . . . . . . . 5.9 Infra-red duality . . . . . . . . . . . . . . . . 5.10 Superconformal invariance in d = 4 . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

180 182 185 185 186 190 190 191 192 194 202 207 210 210 211 213 216 226 229 233 235 237

6 Comments on vacuum energies in scale invariant theories 7 Supersymmetric gauge theories and string theory 7.1 Branes in string theory . . . . . . . . . . . . . . . . . . . . . . . . 7.2 Branes in IIA and IIB string theories . . . . . . . . . . . . . . . . 7.3 The eﬀective ﬁeld theory on branes . . . . . . . . . . . . . . . . . 7.4 Eﬀective D = 4 dimensional systems with N = 2 supersymmetry 7.5 An eﬀective D = 4, N = 1, U (NC ) gauge theory with matter . . 7.6 More pieces of information . . . . . . . . . . . . . . . . . . . . . . 7.7 Obtaining the dual ﬁeld theory . . . . . . . . . . . . . . . . . . . 7.8 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . 8 Final remarks

. . . . . . . .

**Lecture 3. An Introduction to Duality Symmetries in String Theory by A. Sen
**

1 Introduction 2 A brief review of perturbative string 2.1 The spectrum . . . . . . . . . . . . . 2.2 Interactions . . . . . . . . . . . . . . 2.3 Compactiﬁcation . . . . . . . . . . . theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

241

243 245 246 251 254

3 Notion of duality symmetries in string theory 255 3.1 Duality symmetries: Deﬁnition and examples . . . . . . . . . . . . 255 3.2 Testing duality conjectures . . . . . . . . . . . . . . . . . . . . . . 258

xxviii

4 Analysis of low energy eﬀective ﬁeld theory 4.1 Type I–SO(32) heterotic duality in D = 10 . 4.2 Self-duality of heterotic string theory on T 6 . 4.3 Duality between heterotic on T 4 and type IIA 4.4 SL(2, Z) self-duality of type IIB in D = 10 . 4.5 Other examples . . . . . . . . . . . . . . . . .

. . . . on . . . .

. . . . K3 . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

260 261 264 268 272 274

5 Precision test of duality: Spectrum of BPS states 5.1 SL(2, Z) S-duality in heterotic on T 6 and multi-monopole spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2 SL(2, Z) duality in type IIB on S 1 and D-branes . . . . . 5.3 Massless solitons and tensionless strings . . . . . . . . . . 6 Interrelation between diﬀerent duality conjectures 6.1 Combining non-perturbative and T -dualities . . . . . 6.2 Duality of dualities . . . . . . . . . . . . . . . . . . . 6.3 Fiberwise duality transformation . . . . . . . . . . . 6.4 Recovering higher dimensional dualities from lower dimensional ones . . . . . . . . . . . . . .

276 moduli . . . . . 280 . . . . . 286 . . . . . 296

299 . . . . . . . . 299 . . . . . . . . 299 . . . . . . . . 301 . . . . . . . . 304

7 Duality in theories with less than sixteen supersymmetry generators 305 7.1 Construction of a dual pair of theories with eight supercharges . . 306 7.2 Test of duality conjectures involving theories with eight supercharges . . . . . . . . . . . . . . . . . . . . . . . . 309 8 M-theory 312 8.1 M-theory in eleven dimensions . . . . . . . . . . . . . . . . . . . . 312 8.2 Compactiﬁcation of M-theory . . . . . . . . . . . . . . . . . . . . . 315

**Lecture 4. Les Houches Lectures on Large N Field Theories and Gravity by J. Maldacena
**

1 General introduction

323

325

2 The correspondence 330 2.1 The ﬁeld ↔ operator correspondence . . . . . . . . . . . . . . . . . 336 2.2 Holography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 338 3 Tests of the /CFT correspondence 3.1 The spectrum of chiral primary operators . . . . . . 3.1.1 The ﬁeld theory spectrum . . . . . . . . . . . 3.1.2 The string theory spectrum and the matching 3.2 Matching of correlation functions and anomalies . .

Ë

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

341 342 342 348 351

xxix

4 Correlation functions 353 4.1 Two-point functions . . . . . . . . . . . . . . . . . . . . . . . . . . 355 4.2 Three-point functions . . . . . . . . . . . . . . . . . . . . . . . . . 359 5 Wilson loops 362 5.1 Wilson loops and minimum surfaces . . . . . . . . . . . . . . . . . 363 5.2 Other branes ending on the boundary . . . . . . . . . . . . . . . . 368 6 Theories at ﬁnite temperature 369 6.1 Construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 369 6.2 Thermal phase transition . . . . . . . . . . . . . . . . . . . . . . . 372

**Lecture 5. D-Branes on the Conifold and N = 1 Gauge/Gravity Dualities by I.R. Klebanov, C.P. Herzog and P. Ouyang
**

1 Introduction

383

385

2 D3-branes on the conifold 388 2.1 Dimensions of chiral operators . . . . . . . . . . . . . . . . . . . . 391 2.2 Wrapped D3-branes as “dibaryons” . . . . . . . . . . . . . . . . . . 393 2.3 Other ways of wrapping D-branes over cycles of T 1,1 . . . . . . . . 394 3 The RG cascade 397 3.1 Matching of the β-functions . . . . . . . . . . . . . . . . . . . . . . 400 4 The chiral anomaly 402 4.1 The anomaly as a classical eﬀect in supergravity . . . . . . . . . . 403 4.2 The anomaly as spontaneous symmetry breaking in AdS5 . . . . . 405 5 Deformation of (KS) the conifold 410 5.1 The ﬁrst-order equations and their solution . . . . . . . . . . . . . 412 5.2 SO(4) invariant expressions for the 3-forms . . . . . . . . . . . . . 413 6 Infrared physics 6.1 Dimensional transmutation and conﬁnement . . . . . . . . . . . . . 6.2 Tensions of the q-strings . . . . . . . . . . . . . . . . . . . . . . . . 6.3 Chiral symmetry breaking and gluino condensation . . . . . . . . . 414 414 416 418

**Lecture 6. De Sitter Space by M. Spradlin, A. Strominger and A. Volovich
**

1 Introduction

423

425

xxx

2 Classical geometry of de Sitter space 427 2.1 Coordinate systems and Penrose diagrams . . . . . . . . . . . . . . 428 2.2 Schwarzschild-de Sitter . . . . . . . . . . . . . . . . . . . . . . . . . 435 2.3 Geodesics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 436 3 Quantum ﬁeld theory on de Sitter space 437 3.1 Green functions and vacua . . . . . . . . . . . . . . . . . . . . . . . 437 3.2 Temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 440 3.3 Entropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 443 4 Quantum gravity in de Sitter space 446 4.1 Asymptotic symmetries . . . . . . . . . . . . . . . . . . . . . . . . 446 4.2 De Sitter boundary conditions and the conformal group . . . . . . 447 A Calculation of the Brown-York stress tensor 451

**Lecture 7. String Compactiﬁcation with N = 1 Supersymmetry by M.R. Douglas
**

1 Introduction and the question of vacuum selection 2 From ten dimensional geometry to four dimensional eﬀective ﬁeld theory 3 D-branes with stringy corrections 4 Quantum corrections 5 Towards the low energy theory of everything

455

457

461 466 470 472

**Lecture 8. Lectures on Open Strings, and Noncommutative Gauge Theories by N.A. Nekrasov
**

1 Introduction 2 Background independence 2.1 Dolan-Nappi solutions . . . . . . . . . . . . . . . . . . . . . . . . . 2.2 Intersecting branes . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3 T -duality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

477

479 480 483 485 485

xxxi

3 BPS algebras 486 3.1 Noncommutative U (1) instantons . . . . . . . . . . . . . . . . . . . 486 3.2 Higher dimensional instantons . . . . . . . . . . . . . . . . . . . . . 489 4 Fermions in the Y background 490 4.1 Fermions in the instanton background . . . . . . . . . . . . . . . . 490 4.2 The Dirac ﬁeld in the monopole background . . . . . . . . . . . . . 491 5 Non-trivial backgrounds 492 5.1 Example: Γ = Z2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 493 6 Conclusions and outlook 494

**Lecture 9. Condensates Near the Argyres-Douglas Point in SU(2) Gauge Theory with Broken N = 2 Supersymmetry by A. Gorsky
**

1 Introduction 2 Matter and gaugino condensates

497

499 501

3 Dyon condensates 504 3.1 Monopole condensate . . . . . . . . . . . . . . . . . . . . . . . . . . 505 3.2 Charge and dyon condensates . . . . . . . . . . . . . . . . . . . . . 506 4 The Argyres-Douglas point: How well the theory is deﬁned 5 Conclusions 508 509

**Seminar 1. Quantum Field Theory with Extra Dimensions by L. Baulieu
**

1 Introduction 2 The ﬁelds 3 The action at level

513

515 516

Ò

518

3. . . . . . . . .2 Instantons and large N . . . . . . . . . . . . . . . . .5 Further insights . . . . . . . 3. . . . . 3 Special holonomy spaces. 3. . . . . . 3. . . . . .3 Resolved D2-brane . . . . . Gibbons. . . . . . . . . . . . . . . . . . Ferrari 1 Introduction 2 Many paths to the gauge/string 2. . . . . . . . .4 Non-critical strings . . . . . . . . . . . . . . . . . . . . . . . 2. . .2 Old G2 holonomy metrics and their harmonic forms . 2. . . . . . . .1 The old metric and harmonic 4-forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 547 549 550 550 551 552 554 559 562 564 566 569 571 . . . . . .4 Applications: Resolved M2-branes and D2-branes . . . . . . . . . . . .1 Stenzel metric . . . . . . . . . . . . . . . . . . 3. . .2 Harmonic middle-dimension (p. . . .1 Harmonic forms for the Stenzel metric . . 523 525 527 527 528 529 529 530 531 531 532 533 533 534 534 534 535 536 Ú . . . L¨ and C. . . . . . . . . . .1.2 Large N . . . . . . . . . . q) forms . . . . . . . . . . . 3. . G. . 2. . . . . . . . . duality . . . . . . . Pope c u 1 Introduction transgression 2 Resolution 2. . . . . . . . . . 4 New G2 holonomy metrics 4. . . . . . . . . . . . . .N. . 2. . . . . .1 Motivation . . . . . . . .2. . . . . . 3. . . . . . . .2 Resolved M2-brane . . . .W. . . . .3. . . . . . . . . . . . . .2 The new Spin(7) holonomy metric . . . . . . . . . . . .4 Exact results in 4D string theory . .1. . . . . . . . . . . . . . . . .xxxii Seminar 2. . 2. . . . . . . . . . . . . . .2. . . . . . . . . . . . . . . . . . . . . . . . 3. .1 Conﬁnement . . . 3.1 Four dimensional CFTs as Kazakov critical points 3.1 Resolved cones over S 2 × S 4 and S 2 × CP2 . . . . . . . . .1 Classiﬁcation of G2 holonomy spaces with S 3 × S 3 orbits 5 Conclusions and open avenues 537 . . . H. .2 Resolved cone over S 3 × S 3 . . . . .3 A toy model example . . . . . . 3 Four dimensional non-critical strings 3. . . . . . Four Dimensional Non-Critical Strings by F. .4 Other examples . . . . . . . . . .3. . . . . . . . harmonic forms and resolved branes 3. 3. . . . . . . . . . . . . . . . . . . . . . . . Cvetiˇ. . 537 541 Seminar 3. . . . . . . . 3.3 New Spin(7) holonomy metrics and their harmonic forms . . . . . . . . . . . Special Holonomy Spaces and M-Theory by M. . . . . . 2.3 D-branes . . . . 3.

. .1 Free OWLs . . . . . . . . . . 607 Seminar 6. . . . . .5. 595 4 Free and interacting OWLs 597 4. 600 5 Closed strings out of open strings 604 5. . . . . . . 604 5. . 605 5. . . . . . .3 Closed strings as OWLs . . . . .5. . . . . . . . . . . Open Wilson Lines and UV-IR Duality by Soo-Jong Rey 1 Introduction and conjectures 2 Flying noncommutative dipole 587 589 591 3 Open wilson lines: How and why? 593 3. . . . . . . . . . . Exact Answers to Approximate Questions – Noncommutative Dipoles. . . . . . . . . . . . . . . . . . U-Opportunities: Why is Ten Equal to Ten? by B. . . . . . . . . . . . . . . . . . . . . . . . . . . 572 572 4 Open problems Seminar 4. . . . . . . . . . . . . . . . . .1 3.2 Witten’s w -product is Moyal’s m -product . . . . . . . . . . . . . . . . . . . . . Sagnotti 1 Broken supersymmetry and type-0 models 611 613 .2 Generalized star products . .2 Interacting OWLs . . .1 Open Wilson lines . Open-String Models with Broken Supersymmetry by A. . . . 593 3. . . . . . . Julia 1 Classiﬁcations 2 Real forms of Lie algebras 3 Chaos controlled by symmetry 575 577 581 583 Seminar 5. .2 Full proofs . . . . . . . . . . . . . 598 4. . . . . . . . . . .1 Open strings as miniature dipoles . . 571 Non-perturbative non-Borel summable partition functions . . . . . . . . . . . . .xxxiii 3. . .

Shatashvili Seminar 8.L. On a Field Theory of Open Strings. Exceptional Magic by S. Tachyon Condensation and Closed Strings by S.L.xxxiv 2 Scherk-Schwarz deformations and brane supersymmetry 3 Brane supersymmetry breaking 618 620 Seminar 7. Shatashvili 1 G2 2 627 643 648 657 661 ËÔ Ò(7) 3 Topological twist .

Piscataway. The State University of New Jersey. Cambridge CB3 OWA. . SPhT.K. CEA Saclay. NEKRASOV. CANDELAS. MA 02138. I. Rutgers. M. DOUGLAS. France A.A. Oxford OX1 3LB . Wilberforce Road. RABINOVICI. 24 rue Lhomond. Mathematical Institute. Giles. Israel A. Princeton. Switzerland M. N.S. DAMPT. Jhusi. 24-29 St. 91440 Bures-sur-Yvette. France F. J. Institut de Physique. Jefferson Physical Cambridge.A.ORGANIZERS C. India of Laboratory. BACHAS. BILAL. 35 route de Chartres. Université de Neuchâtel. Princeton University. 1 rue Breguet. U.S. SEN. Chhatnag Road.K. U.H. I. University. NJ 08854-8019. MALDACENA. GREEN. France LECTURERS P. 75231 Paris. Racah Institute 91904 Jerusalem. Physics. Institute. U.S. U. Laboratoire de Physique Théorique. KLEBANOV. NJ 08544. E. Oxford University.A. U.S. Harvard Hebrew University. DAVID.E. 2000 Neuchâtel. Harish-Chandra Research 211019 Allahabad. Department of Physics & Astronomy. 91191 Gif-sur-Yvette. Joseph Henry Laboratories. ENS.

CVETIC. FREEDMAN. 75231 Paris. Harvard University. University Roma II. Italy S. Jefferson Physical Cambridge. Tour 16. Université Pierre et Marie 4 place Jussieu. STROMINGER. Center for Theoretical 151-747 Seoul. PA 19104. France Curie. GORSKY. LPT/ENS. University D. B. 24 rue Lhomond. Institute for Theoretical Physics. The Netherlands SEMINAR SPEAKERS L. CERN Theory Division.S. REY.xii A. Center for Theoretical Physics.A. U. Via della Ricerca Scientifica 1. 1211 Genova 23. Russia B. A. Cheremushkinskaya 25. Seoul National University. Switzerland S. 35P4 CE Utrecht.S. Utrecht University. Laboratory. LPTHE. CT 06520. MAYR. Corea Physics. Department of Physics. Tor Vergata. 00133 Roma. U. U. A. 117259 Moscow.S. Department of Physics. Yale University. Spinoza Institute. MIT. MA 02138. JULIA. New Haven. Department of Physics and Astronomy. Cambridge. SAGNOTTI.S.A. Philadelphia. Pennsylvania. . France P. U. ITEP. 75231 Paris Cedex 05. DE WIT. SHATASHVILI.A. B. MA 02139. BAULIEU.A. of M.

Switzerland N. D’APPOLLONIO. 110. Université Pierre et Marie Curie. Germany L. Physics Faculty. 117259 Moscow.S. Israel V. Sezione INFN. Tour 16. BELOV. NJ 08540. CARLEVARO. Russia R. LPTHE.A. Cheremushkinskaya 25. Princeton. Theoretical Physics department. . DOLGUSHEV. Joseph Henry Laboratories. Humboldt University. Steklov Mathematical Moscow 117966. Moscow State University. 50125 Firenze. 10115 Berlin. Cheremushkinskaya ul 25. Dipartimento di Fisica. Racah Institute of Physics.PARTICIPANTS D. B. GSP-1. Jerusalem 91904. D. COUCHOUD. 4 place Jussieu. DYMARSKY.S. CA 93106. ABANIN. Santa Barbara. 8. Physik QFT. Vorobevy Gory. France G. B. Université de Neuchâtel.A. Département de Physique Théorique. ITEP. Department of Physics. Moscow 119899. Princeton University. Russia A. BENA. 75252 Paris Cedex 05. ALEXANDROV. FERRARI. The Netherlands Instituut voor Theoretische Fysica. Tomsk 63050. Russia F. I. A. Russia D. 117259 Moscow. U. Hebrew University. L. BERMAN. Russia Institute. 1018 XE Amsterdam. University of California. 36. DUIVENVOORDEN. U. Invalidenstr.go Fermi 2. ITEP. Tomsk State University. Lenin Av. BRAUN. 2002 Neuchâtel. Institut de Physique. Italy V. University Valckenierstraat 65. Rue Breguet 1. Physics Department. Gubkin St. of Amsterdam. Jadwin Hall.

CA 94305. NJ 08544. Mathematical Institute. KONECHNY. FLOREA. Göteborg University. Department of Theoretical Physics.S. Physics Department. Mumbai 400005. Harvard University. Cambridge. Department of Physics. MASINA. U. India . Colaba. S. HERZOG. Giles. HAACK. CA 94720. U. 06108 Halle. Vorobevy Gory.S. CPHT. bâtiment 774.S. M. Japan V. France S. Friedmann Bach Platz 6. V. LOW. U. D. University of California Berkeley. Theoretical Physics Group. KASTE.S. Jefferson Physical Laboratories. Kyoto University. KLEVTSOV. MALYSHEV. University of Oxford.A. I. Stanford. Princeton. Fachbereich Physik. 382 via Pueblo Mall. Moscow 119899. Orme des Merisiers.S. S. France Théorique Luminy.A. 382 Via Pueblo Mall. 91191 Gif-sur-Yvette. Quantum Statistics and Quantum Field Theory Department.A. SPhT. HUBENY. Department of Theoretical Physics. Moscow State University. U. Case 907. Princeton University. Mail Stop 50A-501 4BNL. Cambridge.K. Martin Luther Universität Wittenberg. Moscow 119899. 24-29 St. Homi Bhabha Road. KitashirakawaOiwakecho. Tata Institute of Fundamental Research. Stanford University. NEMANI. Harvard University. Berkeley. Physics Faculty. Vorobevy Gory. Kyoto 606-8502. MA 02138. Department of Physics. CEA Saclay. U.A.S. Germany P. CA 94305-4060. Oxford OX1 3LB. MORIYAMA. École Polytechnique. 75005 Paris. France C. Stanford. HENRY. McALLISTER. Stanford University. Sakyo-ku. Sweden S.S. 91128 Palaiseau. 24 rue Lhomond. Russia A. Department of Physics. U. France L. Department of Physics. ENS. MA 02138. Chalmers University of Technology. U. P.A. Moscow State University. Russia I. GUKOV. U. 41296 Gothenburg. GRAN. LPT. Theoretical Physics Department. Physics Faculty. Centre de Physique 13288 Marseille Cedex 9.xiv B.A. GURRIERI.

D. J. NJ 08854-8019. U. Washington Road. SPhT. 4 Washington Place. NOGRADI. Departamento De Fisica Teorica C-XI. Italy G. 452-48 Cal Tech. SARAIKIN. 91191 Gif-sur-Yvette. Russia S. RICCIONI. 00133 Roma. SMEDBÄCK. V. University of Leiden. The State University of New Jersey. NITTI. SCHÄFER-NAMEKI. Centre for Mathematical Sciences. Fine Hall. F. SISSA. Department of Theoretical Physics.A.O. NY 10003. Dipartimento di Fisica. U. M. France Théorique. 136 Frelinghuisen Road.O. Wilberforce Road. via della Ricerca Scientifica 1. Box 2208.S. Cheremushkinskaya 25. Naukova St. 117259 Moscow. 28049 Madrid. RIBAULT. V. U. RYCHKOV. 1200 East California Blvd. Dnepropetrovsk National University. Greece Crete. Spain S.S. PESTUN. Nijenborgh 4. CA 91125.xv F. K.A. Theoretical Physics Department. PARK.S. Randall Laboratory of Physics. Russia R. SERVANT.A. New York. Michigan Center for Theoretical Physics. The Netherlands C. B. 34014 Trieste. Via Beirut 4. Italy D. Rutgers. RABADAN.A. Box 803. ROEST. NJ 08544. Universita di Roma “Tor Vergata”. 751 08 Uppsala.A. University of Cambridge. Box 9506. Institute for Theoretical Physics. New York University. France A. École Polytechnique. Mathematics Department. Ann Arbor. P. Dnepropetrovsk 49050. The University of Michigan. ROMELSBERGER. Princeton University. Cheremushkinskaya 25.K. 2300 RA Leiden. Centre de Physique 91128 Palaiseau. B. 117259 Moscow. SHCHERBAKOV. Institute Lorentz. Department of Physics & Astronomy. U. California Institute of Technology. ITEP. Piscataway. Princeton. Cambridge CB3 0WA.S. Pasadena. CEA Saclay.. DAMTP. Physics Department. Universidad Autonoma de Madrid. Ukraine M. ITEP. PANDO ZAYAS. MI 48109-1120. Uppsala University. 9747 AG Groningen. G. 13. PANOTOPOULOS. U. The Netherlands L. U. University of P. Department of Physics. Sweden . Heraklion.S. SCHNABL.

A. Harvard University. ULANOV. 8093 Zurich. WALCHER. B. U. ZOTOV. Kiev. 117259 Moscow. Seoul 151-742. 13. ETH. S. YEE. Zurich. Russia A. ITEP. Cambridge. St. Cambridge CB3 OWA. Ul B. Institute for Theoretical Physics. U. 14 B Metrologicheskaya St. School of Physics. Seoul National University. MA 02138. ITEP. Cambridge. SPRADLIN. J. VOLOVICH. Naukova St.xvi A. 17A Oxford Street.S. 117259 Moscow. Department of Applied Mathematics and Theoretical Physics. C. Russia . Harvard University. VASSILIEV. Dnepropetrovsk National University. Cheremushkinskaya 25. Ukraine M..A. Theoretical Physics Department.-T. U. STAHN. Physics MA 02138.K. SOLOVYOV. Switzerland J. MIPT.S. Dnepropetrovsk 49050. Department of Physics. Cheremushkinskaya 25. South Korea A. Ukraine D. Department. Institute for Theoretical Physics.

LECTURE 1 SUPERGRAVITY B. Utrecht University. The Netherlands . DE WIT Institute for Theoretical Physics & Spinoza Institute.

. . . .2 On ungauged maximal supergravity Lagrangians 5.5 Frames and ﬁeld redeﬁnitions . . .4 Dimensional reduction and hidden symmetries . . . . . . . 2. . . . . . . . the T -tensor . . . . . . . . . . . . . . . . . 5. . . . . . . . . . . . . . . . . . . . . . . . . . . . 119 7. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 The superalgebras OSp(N |4) . . . . . . . . . . . . . . . . . . . . . . . . . 4. . . . . .2 Maximal supersymmetry and supergravity . . . 97 6. . . . . . . . .2 Massless supermultiplets . . . 2. .6 Kaluza–Klein states and BPS-extended supergravity 3. . . . . . . 5. . . . . . . . . . .1 The superconformal algebra . . . . . . . . . .1 Nonlinearly realized symmetries . . . . . . . . 5 Gauged maximal supergravity in 4 and 5 dimensions 5. . . . . . . . .3 Massive supermultiplets . . . . . . . . . . . .1 The Poincar´ supersymmetry algebra . . . . . . . . . . . . . 126 . . . . . . . . . . . . 3. . .2 Unitary representations of the anti-de Sitter algebra .5 On spinors and the R-symmetry group HR 3 4 5 7 17 19 22 28 28 32 35 38 43 46 53 55 56 61 66 69 70 72 74 79 84 . . . . . . e 2. . . 4. . . . . . . 122 7. . .2 Geometrical quantities . . . . . . . 2.Contents 1 Introduction 2 Supersymmetry in various dimensions 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Supergravity 3.4 Gauging maximal supergravity. . . . . . . . . . . . . . . . . . . . . 3. .4 Central charges and multiplet shortening . . . . . . 3. . .7 Nonmaximal supersymmetry: Q = 16 . . . . . . .1 On E7(7) /SU (8) and E6(6) /U Sp(8) cosets . . . . . . . . . . . . . . . . . . . . . . . . . 3. space . . . .2 Superconformal gauge theory and supergravity . . . . . . . . . 4 Homogeneous spaces and nonlinear sigma models 4. . .3 D = 11 supergravity . . . . . . . 110 7 Superconformal symmetry 116 7. . . 3. . . . . . . . . . . . .1 Simple supergravity .4 Gauged nonlinear sigma models . . .3 Nonlinear sigma models with homogeneous target 4. . . .3 Matter ﬁelds and currents . . . . . . . . . .3 Electric–magnetic duality and E7(7) . . . 6 Supersymmetry in anti-de Sitter space 89 6. . . . . . . . . . . . . . . . . . . . . . . . . . . . 93 6. . . . . . .1 Anti-de Sitter supersymmetry and masslike terms .

Jan Louis. which will cover a number of basic aspects of supergravity. somewhat hypothetical. u Hermann Nicolai. Looking backwards it becomes clear that there are many reasons why neither superstrings nor supergravity could account I am grateful to Sergio Ferrara. e. The only candidate for such a theory is superstring theory (for some reviews and textbooks. Also in these lectures. called M-theory. String theory is no longer a theory exclusively of strings but includes other extended objects that emerge in the supergravity context as solitonic objects. But nowadays supergravity is primarily viewed as an eﬀective ﬁeld theory describing the low-mass degrees of freedom of a more fundamental underlying theory. Springer-Verlag 2002 . de Wit 1 Introduction Supergravity plays a prominent role in our ideas about the uniﬁcation of fundamental forces beyond the standard model. Soo-Jong Rey. theory. Stefan Vandoren and Toine Van Proeyen for many helpful and stimulating discussions. During its historical development the perspective of supergravity has changed. [2]). The advances made have found their place in many reviews and textbooks (see. Murat G¨ naydin. Mario Trigiante. Although we know a lot about Mtheory.g. and in recent developments of the conceptual basis of quantum ﬁeld theory and quantum gravity. in our understanding of many central features of superstring theory.SUPERGRAVITY B. see. e. Peter van Nieuwenhuizen. Henning Samtleben. many topics will be left untouched. String theory and supergravity in their modern incarnations now represent some of the many faces of M-theory. c EDP Sciences. its underlying principles have only partly been established. Kostas Skenderis.g. [1]). but the subject has grown so much and has so many diﬀerent facets that no comprehensive treatment is available as of today. Francisco Morales. yet another. Ergin Sezgin. or rather. Originally it was envisaged as an elementary ﬁeld theory which should be free of ultraviolet divergencies and thus bring about the long awaited uniﬁcation of gravity with the other fundamental forces in nature.

This is one of the motivations for considering anti-de Sitter supersymmetry and the representations of the anti-de Sitter group in Section 6. To appreciate the implications of supersymmetry. In Section 4 gauged nonlinear sigma models with homogeneous target spaces are introduced. In this section1 we concentrate on supermultiplets of states. primarily restricting ourselves to ﬂat Minkowski spacetimes of dimension D. This section is self-contained. These theories can describe anti-de Sitter ground states which are fully supersymmetric. Section 2 starts with a detailed discussion of supersymmetry and its representations. because of nonrenormalization theorems. Subsequently supergravity theories are introduced in Section 3. This construction is explained in Section 5. It controls the dynamics and. The relevant symmetries in this case form an 1 The material presented in this and the following section is an extension of the second section of [6]. Our hope is that the material presented below will oﬀer a helpful introduction to and will blend in naturally with the material presented in other lectures. paving the way for the construction of gauged supergravity. 2 Supersymmetry in various dimensions An enormous amount of information about supersymmetric theories is contained in the structure of the underlying representations of the supersymmetry algebra (for some references. This school oﬀers a large number of lectures dealing with gravity. mostly concentrating on the maximally supersymmetric cases. Here we should make a distinction between a supermultiplet of ﬁelds which transform irreducibly under the supersymmetry transformations. 3–6]). where the emphasis is on gauged supergravity with 32 supercharges in 4 and 5 spacetime dimensions. gauge theories and string theory from various perspectives.4 Unity from Duality: Gravity. precise predictions can be made in many instances. see [1. often relating strong. We intend to stay within the supergravity perspective and to try and indicate what the possible implications of supersymmetry and supergravity are for these subjects. Section 7 contains a short introduction to superconformal transformations and superconformally invariant theories. Because supersymmetry is such a powerful symmetry it plays a central role in almost all these developments. Gauge Theory and Strings for all the relevant degrees of freedom and we have learned to appreciate that M-theory has many diﬀerent realizations.to weak-coupling regimes. and a supermultiplet of states described by a supersymmetric theory. . but it is of course related to the discussion in Section 6 on anti-de Sitter representations as well as to the adS/CFT correspondence.

The spinors are then Dirac spinors. +). all charges that commute with the supercharges. e Observe that the issue of central charges is diﬀerent when not in ﬂat space. for example. de Wit: Supergravity 5 extension of the Poincar´ transformations. are called “central charges”. When this dimension is odd. 2. However. the generators of the Lorentz group. many of the concepts that we introduce will also play a role in the discussion of other superalgebras. often called central charges2. corresponding to spacetimes of particular dimensions.with negative-chirality spinors. For the moment we ignore these additional charges. such as the antide Sitter (or conformal) superalgebras. in the context of the anti-de Sitter superalgebra (discussed in Sect. e transforming as spinors under the Lorentz group. All spinors in this sequence can be restricted to Majorana spinors.B. For a recent practical introduction to superalgebras. There are other relevant superalgebras. We adhere to this nomenclature. and possibly additional generators that commute with the supercharges. Qβ } = −2iPµ (Γµ )αβ . see [7]. +. These cases constitute the ﬁrst sequence. For the remaining two sequences. (2. there exist ﬁve diﬀerent sequences of spinors. charge conjugation relates positive. where the number of dimensions is a multiple of 4. but not necessarily with all the generators of the Poincar´ algebra. Usually. it is possible in certain cases to have Majorana spinors. As shown. . The ﬁrst step is therefore to determine Q for any given number of spacetime dimensions D. Now there are again two possibilities. which consist of translations and e Lorentz transformations. These will be encountered in due course. In even dimension one may distinguish three sequences. The second one corresponds to those odd dimensions where Majorana spinors do not exist. In the ﬁrst one. ¯ {Qα . 6). such as the supersymmetric extensions of the antide Sitter (or the conformal) algebras. The most important anti-commutation relation of the super-Poincar´ e algebra is the one of two supercharges. The result is summarized in Table 1. as can be seen. depending on whether Majorana spinors 2 The terminology adopted in the literature is not always very precise.1) where we suppressed the central charges.1 The Poincar´ supersymmetry algebra e The generators of the super-Poincar´ algebra comprise the supercharges. · · · . the energy and momentum operators. The size of a supermultiplet depends exponentially on the number of independent supercharge components Q. charge conjugation preserves the chirality of the spinor. 1) with Minkowskian metric ηµν = diag(−. Here Γµ are the gamma matrices that generate the Cliﬀord algebra C(D − 1.

Majorana spinors exist and the charges can be restricted to so-called Majorana-Weyl spinors. Weyl (W) and Majorana-Weyl (MW). in the context of the Euclidean rest-frame superalgebra for massive representations or for the anti-de Sitter superalgebra. The third column speciﬁes the group HR for N -extended supersymmetry. one can consider combinations of N+ positiveand N− negative-chirality spinors. For Weyl charges. In addition we include the number of . one for each chiral sector. The supercharges in ﬂat Minkowski spacetimes of dimension D. comprise the fourth sequence. 10. mod 8 Qirr 2(D−1)/2 2(D+1)/2 2D/2 2D/2 2D/2−1 HR SO(N ) U Sp(2N ) U (N ) U Sp(2N+ ) × U Sp(2N− ) SO(N+ ) × SO(N− ) type M D M W MW can exist or not. Another way to present some of the results above. mod 8 2. corresponding to D = 6 mod 8. is thus deﬁned as the largest subgroup of the automorphism group of the supersymmetry algebra that commutes with the Lorentz group. mod 8 6. Here we have assumed that HR is compact so that it preserves a positive-deﬁnite metric. but this is by no means necessary. In all these cases there exists a group HR of rotations of the spinors which commute with the Lorentz group and leave the supersymmetry algebra invariant. For the last sequence with D = 2 mod 8. There are other versions of the R-symmetry group HR which play a role. Qirr speciﬁes the real dimension of an irreducible spinor in a Ddimensional Minkowski spacetime. is shown in Table 2. often referred to as the “R-symmetry” group. 9. It is often realized as a manifest invariance group of a supersymmetric ﬁeld theory. Gauge Theory and Strings Table 1. 8. mod 8 5. In the second column. Dirac (D). 11. One can consider extended supersymmetry. mod 8 4. we allow N± charges of opposite chirality. where the spinor charges transform reducibly under the Lorentz group and comprise N irreducible spinors. This group. D 3. The cases where we cannot have Majorana spinors. In Table 1 we have listed the corresponding HR groups for N irreducible spinor charges. for instance. In the latter two sequences of spinor charges shown in Table 1. The fourth column denotes the type of spinors: Majorana (M). Here we list the real dimension of an irreducible spinor charge and the corresponding spacetime dimension.6 Unity from Duality: Gravity. 7. acting on N -fold reducible spinor charges. deﬁned in the text. Those will be discussed later in this section. so that HR decomposes into the product of two such groups.

5 =4 =3 shortest supermultiplet 128 + 128 8+8 4+4 2+2 1+1 states of the shortest3 supermultiplet of massless states. Qirr 32 16 8 4 2 D D D D D D = 11 = 10. (2. 9. 2.1) as (using ¯ Q = iQ† Γ0 ).4) 3 By the shortest multiplet. de Wit: Supergravity 7 Table 2.e. and ˜ ⊥ of the product of all D − 2 gamma matrices in the transverse directions Γ (i. for massless representations.5. Q† } = 2 P 0 1 + ΓD Γ⊥ . written as a sum of bosonic and fermionic states. perpendicular to P ). {Qα . P ) the right-hand side is proportional to a projection operator (1 + Γ Γ0 )/2. Simple supersymmetry in various dimensions. We return to a more general discussion of the R-symmetry groups and their consequences in Section 2.1) has therefore zero eigenvalues. 8. we mean the multiplet with the helicities of the states as low as possible. We present the dimension of the irreducible spinor charge with 2 ≤ Qirr ≤ 32 and the corresponding spacetime dimensions D.3) β αβ ˜ Here ΓD consists of the product of all D independent gamma matrices. let us rewrite (2. Γ⊥ ] = 0 . which. To exhibit this more explicitly. The matrix Pµ Γµ on the righthand side of (2. The supersymmetry anti-commutator can then be written as ˜ ˜ {Qα . This is usually (one of) the smallest possible supermultiplet(s).2) β For light-like P µ = (P 0 .B. we may consider the states at arbitrary but ﬁxed momentum Pµ . with phase factors such that ˜ ˜ (ΓD )2 = (Γ⊥ )2 = 1 .. satisﬁes P 2 = 0. The third column represents the number of bosonic + fermionic massless states for the shortest supermultiplet. In a positive-deﬁnite Hilbert space some (linear combinations) of the supercharges must therefore vanish. (2. . / (2. Here Γ is the gamma matrix along the spatial momentum P of the states.2 Massless supermultiplets Because the momentum operators Pµ commute with the supercharges. ˜ ˜ [ΓD . Q† } = 2 (P Γ0 )αβ . 7 = 6.

24. In D = 6 there are three theories with Q = 32 and one with Q = 24. one corresponding to the bosons and the other one to the fermions.3) simpliﬁes and shows that the 16 nonvanishing spinor charges transform according to a single spinor representation of the helicity group SO(9).1 D = 11 supermultipets In 11 dimensions we are dealing with 32 independent real supercharges. 12. Pure supergravity theories with spacetime dimension 4 ≤ D ≤ 11 can exist with Q = 32.3) is proportional to a projection operator. Gauge Theory and Strings This shows that the right-hand side of (2. The derivation for D = 4 is presented in many places (see. 2. it turns out that the dimension of the shortest supermultiplet is not just equal to 2Qirr /4 . 10 and 6 spacetime dimensions. one considers 4 In D = 4 there exist theories with Q = 12. We shall demonstrate this below in three relevant cases. 20. On the other hand. The two chiral spinor subspaces correspond to the bosonic and fermionic states. whereas the other ones generate a Cliﬀord algebra. 16. this is so because the representation is complex. In odd-dimensional spacetimes irreducible spinors are subject to the eigenvalue ˜ condition ΓD = ±1. corresponding to D = 11. when regarding the 16 spinor charges as gamma matrices. 4. So far these supergravities have played no role in string theory. 8 and 4 supersymmetries4 . the dimensions are shown in Table 2. Some of these theories will be discussed later in more detail (in particular supergravity in D = 11 and 10 spacetime dimensions). alternatively. For D = 6. respectively. Consequently. which decomposes into two chiral subspaces. the representation space of the charges decomposes into the two chiral spinor representations of SO(Q/2). For the massless multiplets. . Depending on the number of spacetime dimensions. For D = 3 we refer to [8]. half the spinors must vanish on physical states. 4 the representation is twice as big because it must also accommodate fermion number (or. When confronting these results with the last column in Table 2. it follows that the representation space constitutes the spinor representation of SO(16). For D = 3.8 Unity from Duality: Gravity. for instance [1. as one might naively expect. which projects out half of the spinor space. For a more recent discussion. Denoting the real dimension of the supercharges by Q. because it must be CPT self-conjugate). in D = 5 there exists a theory with Q = 24 [9]. 20 and 24. To determine the helicity content of the bosonic and fermionic states. many supergravity theories exist. Therefore (2. 5]). Bigger supermultiplets can be obtained by combining irreducible multiplets by requiring them to transform nontrivially under the Lorentz group.2. see [10].

B. whose irreducible representation space corresponds to the bosonic and fermionic states. denoted as the 8v and 8c representations. and 350 . let us ﬁrst discuss some other properties of SO(8) representations. 350 for 35v . de Wit: Supergravity 9 the embedding of the SO(9) spinor representation in the SO(16) vector representation. These multiplets will not be considered here. 2. Bigger supermultiplets consist of multiples of 256 states. It then turns out that one of the 128 representations branches into helicity representations according to 128 → 44 + 84. the nonvanishing charges transform in a chiral spinor representation of the SO(8) helicity group. nonextended) supersymmetry with 16 charges. respectively. ˜ The latter means that they are eigenspinors of ΓD . is to investigate how they decompose 5 The representations can be characterized according to the four diﬀerent conjugacy classes of the SO(8) weight vectors.2. we assign the bosonic and fermionic states. One is the representation to which we assign the supercharges. With the exception of certain representations. 35c . According to (2. for instance [12])5 . when we have simple (i. 350 . The graviton states transform in the 44. such as the adjoint and the singlet representation. The above states comprise precisely the massless states corresponding to D = 11 supergravity [11]. the smallest massive supermultiplet comprises 32 768 + 32 768 states. each transforming according to a chiral spinor representation. while the second one transforms irreducibly according to the 128 representation of the helicity group. For instance. without central charges. respectively. Hence we are dealing with three 8-dimensional representations of SO(8).2 D = 10 supermultiplets In 10 dimensions the supercharges are both Majorana and Weyl spinors. s and c.e. the antisymmetric tensor states in the 84 and the gravitini states in the 128 representation of SO(9). They are traditionally distinguished by labels s.3). Before constructing the supermultiplets that are relevant for D = 10 supergravity. The smallest massless supermultiplet has now been constructed with 8 bosonic and 8 fermionic states and corresponds to the vector multiplet of supersymmetric Yang-Mills theory in 10 dimensions [13]. 35s . With 8 nonvanishing supercharges we are dealing with an 8-dimensional Cliﬀord algebra. v and c (see. The fact that SO(8) representations appear in a three-fold variety is known as triality. In this context one uses the notation 10 . to the other two. the three types of representation are inequivalent. which we will denote by 8s . One way to distinguish the inequivalent representations.. v. . denoted by 0. which are inequivalent. 280 . which is a characteristic property of the group SO(8).

= 8s + 56s . with one of the 8-dimensional representations. 35v −→ 1 + 7 + 27 . Massless N = 1 supermultiplets in D = 10 spacetime dimensions containing 8 + 8 or 64 + 64 bosonic and fermionic degrees of freedom. Multiplying 8v with 8v + 8c yields 8v × 8v bosonic and 8v × 8c fermionic states. Using the multiplication rules for SO(8) representations. 56v 56c. 1 + 28 + 35c . This supermultiplet contains the representation 35v . Therefore this supermultiplet will be called the graviton multiplet.and 56-dimensional representations are 28 −→ 7 + 21 . Each of the 8-dimensional representations leaves a diﬀerent SO(7) subgroup of SO(8) invariant. supermultiplet vector multiplet graviton multiplet gravitino multiplet gravitino multiplet bosons 8v 1 + 28 + 35v 1 + 28 + 35c 8v + 56v fermions 8c 8s + 56s 8s + 56s 8c + 56c into representations of an SO(7) subgroup. 8v × 8s 8s × 8c 8c × 8v = 8c + 56c .10 Unity from Duality: Gravity. which can be associated with the states of the graviton in D = 10 dimensions (the ﬁeld-theoretic identiﬁcation of the various states has been clariﬁed in many places. −→ 8 + 48 . 35. 1 + 28 + 35s . = 8v + 56v . There are thus three diﬀerent possibilities.s −→ 21 + 35 . 35c.s −→ 35 .5) In order to obtain the supersymmetry representations relevant for supergravity we consider tensor products of the smallest supermultiplet consisting of 8v + 8c . Gauge Theory and Strings Table 3. 8c −→ 8 . 8v × 8v 8s × 8s 8c × 8c = = = 1 + 28 + 35v .g. Multiplication . each leading to a 128-dimensional supermultiplet. see e. (2. Corresponding branching rules for the 28-. the Appendix in [6]). (2. where 8 is the spinor representation of SO(7). Under this SO(7) the other two 8-dimensional representations branch into 8s −→ 8 .6) it is straightforward to obtain these new multiplets. and leads to the second supermultiplet shown in Table 3. Therefore there is an SO(7) subgroup under which the 8v representation branches into 8v −→ 7 + 1.

These supermultiplets contain fermions transforming according to the 56s and 56c representations. which can be associated with gravitino states. we interchange the boson and fermion assignments in these products. A second supermultiplet may be viewed as the tensor product of a (8v +8s ) supermultiplet with a second supermultiplet (8v +8c ). By combining a graviton and a gravitino multiplet it is possible to construct an N = 2 supermultiplet of 128 + 128 bosonic and fermionic states. Because the supercharges have the same chirality. According to the construction presented above. This is the multiplet corresponding to IIB supergravity [14]. graviton and gravitino supermultiplets are shown in Table 3. Hence the automorphism group HR is equal to SO(2). where the 1. Consequently.B. We have thus established the existence of two inequivalent gravitino multiplets. there will also be two inequivalent N = 2 supermultiplets containing the states corresponding to a graviton and two gravitini. 28 and 56s representations are degenerate and constitute doublets under this SO(2) group.7) fermions: 8s + 8s + 56s + 56s . since there are two inequivalent gravitino multiplets. 8s . one can perform rotations between these spinor charges which leave the supersymmetry algebra unaﬀected. Multiplication with 8c then leads to 8c × 8c bosonic and 8c × 8v fermionic states. However. except that we will associate the 8c and 8s representations with fermionic quantities (note that these are the representations to which the fermion states of the Yang-Mills multiplet and the supersymmetry charges are assigned). The explicit SO(8) decompositions of the vector. de Wit: Supergravity 11 with 8c or 8s goes in the same fashion. respectively. . Such a multiplet follows if one starts from a supersymmetry algebra based on two MajoranaWeyl spinor charges Q with the same chirality. In this case the supercharges constitute two Majorana-Weyl spinors of opposite chirality. but no graviton states as those transform in the 35v representation. This feature reﬂects itself in the multiplet decomposition. Therefore these two supermultiplets are called gravitino multiplets. whereas multiplication with 8s gives 8s × 8c bosonic and 8s × 8v fermionic states. The states of this multiplet decompose as follows: Chiral N = 2 supermultiplet (IIB) bosons: 1 + 1 + 28 + 28 + 35 + 35 v c (8v + 8c ) × (8v + 8c ) =⇒ (2. one N = 2 supermultiplet may be be viewed as the tensor product of two identical supermultiplets (namely 8v + 8c ).

Gauge Theory and Strings Now the supermultiplet decomposes as follows: Nonchiral N = 2 supermultiplet (IIA) bosons: 1 + 8 + 28 + 35 + 56 v v v (8v + 8s ) × (8v + 8c ) =⇒ (2. the reductions of IIA and IIB supergravity to 9 dimensions. in D = 9 we have a degeneracy of states.12 Unity from Duality: Gravity. Indeed. This leads to precisely the same supermultiplet as the reduction of the states of IIA supergravity. which in D = 9 dimensions comprises the rest-frame rotations.8) fermions: 8s + 8c + 56s + 56c . They correspond to the degrees of freedom described by the metric. The latter follows from the fact that two D = 10 Majorana-Weyl spinors with opposite chirality can be combined into a single D = 11 Majorana spinor. One could also take the states of the IIB supergravity and decompose them into D = 9 massless states. However. 84 and 128 representation of the helicity group SO(9). Therefore the Kaluza-Klein states that one obtains when compactifying the ten-dimensional theory on a circle remain inequivalent for the IIA and IIB theories (see [19] for a discussion of this phenomenon and its . (2. yield the same theory [16–18]. characterized in terms of representations of the helicity group SO(7): −→ 1 1 −→ 1 + 7 8v 44 =⇒ 35v −→ 1 + 7 + 27 84 =⇒ 8s 8c 128 =⇒ 56s 56c 28 56v −→ 7 + 21 −→ 21 + 35 −→ −→ −→ −→ 8 8 8 + 48 8 + 48. respectively. transforming in the 27 and 48 representations of the SO(7) helicity group. the massive states are still characterized in terms of the group SO(8). The formula below summarizes the massless states of IIA supergravity from an 11-dimensional perspective. a 3-rank antisymmetric gauge ﬁeld and the gravitino ﬁeld. We note the presence of graviton and gravitino states. We also show how the 10-dimensional states can subsequently be branched into 9-dimensional states. It can be obtained by a straightforward reduction of D = 11 supergravity. This is the multiplet corresponding to IIA supergravity [15].9) Clearly. The massless states of 11-dimensional supergravity transform according to the 44. associated with the group HR = SO(2).

2. Tables 10 and 11). but almost all information follows from ﬁrst considering the purely chiral case. respectively. . Let us now turn to speciﬁc supermultiplets. For odd values of N+ the representations are complex. the chiral rotations of the spinors can be extended to the group U Sp(2N+). In principle we have N+ positiveand N− negative-chirality charges.B. an identical table can be given for negative-chirality spinors. vector and 6 The content of this table also speciﬁes the shortest massive supermultiplets in four dimensions as well as with the shortest massless multiplets in ﬁve dimensions. The states transform both in the SU+ (2) helicity group and under a U Sp(2N+ ) group. The SU (2) group is then associated with spin or with helicity. Of course.2. In D = 6 dimensions the helicity group SO(4) decomposes into the product of two SU (2) groups: SO(4) ∼ (SU + (2) × SU − (2))/Z2 . for even N+ they can be chosen real. Maximal supergravity will be introduced in Section 3. Because the charge conjugated spinor has the same chirality. Let us recall that the helicity assignments of the states describing gravitons. Likewise N− negative-chirality spinors transform under U Sp(2N− ). Shortest massless supermultiplets of D = 6 N+ -extended chiral supersymmetry. 2. the smallest supermultiplet will only transform under one SU (2) factor of the helicity group. In Table 4 we present the decomposition of the various helicity representations of the smallest supermultiplets based on N+ = 1. The ﬁeld content of the maximal Q = 32 supergravity theories for dimensions 3 ≤ D ≤ 11 will be presented in two tables (cf. gravitini.3 D = 6 supermultiplets In 6 dimensions we have chiral spinors. 3 or 4 supercharges. It turns out that the Q = 32 supergravity multiplets are unique in all spacetime dimensions D > 2. SU+ (2) 5 4 3 2 1 N+ = 1 N+ = 2 N+ = 3 1 6 14 14 (32 + 32)C 1 2 (2 + 2)C 1 4 5 (8 + 8)R N+ = 4 1 8 27 48 42 (128 + 128)R consequences). for N+ chiral spinors. except for D = 10. de Wit: Supergravity 13 Table 4. This feature is already incorporated in Table 1. which are not Majorana. as is shown in Table 46 . When we have supercharges of only one = chirality.

(2. and a ﬁve-plet of spinless states. where the smallest supermultiplet is the (1. 1) supersymmetry. 2) or (2. 2). This yields the vector multiplet. with a vector state. 1) or (1. 1). with the vector state and four scalars. the latter transforming with respect to the (2. Here (m. 2). 2) . 0) supergravity. 2). (3. n) denotes that the dimensionality of the reducible representations of the two SU (2) factors of the helicity group are of dimension m and n. one obtains the states of (1. because gravitini without a graviton are not expected to give rise to a consistent interacting ﬁeld theory. 1) helicity representation gives the (1. 2). namely the one corresponding to (N+ . Multiplying the hypermultiplet with the (2. 3) representations. 3). 0) tensor multiplet. 1) or (1. with the bosonic states decomposing into a selfdual tensor. then corresponds to the (2. there exists also a nonchiral version with 16 supercharges. 0) tensor multiplet. four chiral gravitini and ﬁve selfdual tensors [20]. 0) and (0. 0) supergravity multiplet. The tensor product with the (1. consisting of the graviton. Next consider (N+ . There are supermultiplets with higher SU (2) helicity representations. the selfdual tensors of the tensor and of the supergravity supermultiplet are of opposite selfduality phase. 3). Taking the tensor product of the smallest supermultiplet with the (2. 3) representation. (2. and spinor ﬁelds are (3.14 Unity from Duality: Gravity. see for instance one of the Appendices in [6]. 0) vector multiplet. and (2. as well as the lower-dimensional ones. 3) helicity representation yields the (2. For the derivation of these assignments. The multiplets of this type are shown in Table 5. we insist that it will also contain a single (3. Some of these multiplets are shown in Table 6 and we will discuss them in due course. 0). Again. (3. 3). Observe that the selfdual tensor ﬁelds in the tensor and supergravity supermultiplet are of opposite selfduality phase. and a four-plet of chiral fermions. Hence we have only (3. shown in Table 4. consisting of a complex doublet of spinless states and a chiral spinor. The smallest multiplet. 3) representations. We now elucidate the construction of the supermultiplets listed in Table 5. There are no such multiplets for more than Q = 32 supercharges. N− ) = (2. Of course. N− ) = (1. 0) supersymmetry. In the following we will ﬁrst restrict ourselves to helicities that correspond to at most the three-dimensional representation of either one of the SU (2) factors. The simplest case is (N+ . (2. 1) or (1. 3) or (3. When a supermultiplet contains (3. The smallest multiplet is now given by the tensor product of the supermultiplets with (1. 0) hypermultiplet. (3. 2) helicity representation yields the (1. Gauge Theory and Strings (anti)selfdual tensor gauge ﬁelds. Multiplication with the (1. 3) helicity representation. a spinless state and a doublet of chiral spinors. N− ) = (1. with a selfdual tensor. 3). which contain neither gravitons nor gravitini. a doublet of chiral spinors and a scalar.

5. 3. 1.c. N− ) supersymmetry. 2. Taking the tensor product of the vector multiplet with the (2. 1. 1) + (1. 1. 1. 1) (3. 1. 3. 5. 0) tensor (1. 2) +(1. 0) graviton (2. 2) (3. 1. 2. 1. 5) + (1. 1) +(1. 0) vector (1. 3. 1. 3. 5. 3. 4. 2. 1) (3. 5. 2. 2. 1. 1) + (2. 2. 2. 1. 3. 1) +(2. 2) +(2. 1. 1. 3. 1. 2. 1) + (1. 2. a tensor. n) representations of the helicity group ˜ ˜ SU+ (2) × SU − (2) and (m. 1. 1) (3. a scalar and four vector states. 1. 4) representation of U Sp(2) × U Sp(2). 5. 2. 1. 1. 2. 2) graviton 128 + 128 fermions (2. 1) +(1. 1) +(2. 1. 1) + (1. 4. corresponding to a graviton. 3. 5. 2) +(2. (3. 1. Some relevant D = 6 supermultiplets with (N+ . 1) graviton 64 + 64 (2. 1) (2. 1) + h. 1) + (1. 1) graviton # 4+4 4+4 4+4 12 + 12 8+8 24 + 24 8+8 32 + 32 bosons (1. 3. 4. 4. 2. n. 2. The 32 fermionic states comprise two doublets of chiral gravitini and two chiral spinor doublets. 3. each transforming as a doublet under the corresponding U Sp(2) group. 1) (1. 1. 2) representation of U Sp(2) × U Sp(2). transforming as doublets under the appropriate U Sp(2) group. 2) (3. 1) (2. 2) representation of the helicity group yields the states of the (1. 1. n) are assigned to (m. 1. 1. 3. 2. 5. 0) graviton (1. 1. 1) +(1. 5. 2. 1. 1) +(1. 2. 1) (2. 1. 2. 1. 2. 3. de Wit: Supergravity 15 Table 5. The states (m. 1. 1) (3. 1) +(3. 1) (3. 5) +(1. 5) (2. multiplet (1. 0) hyper (1. 3. 1. 2. m. 1) + (3. 4) + (1. 4. It consists of 32 bosonic states. 3. 1) (2. 1) supergravity multiplet. 2.B. . There are two doublets of chiral fermions with opposite chirality. 1) vector (1. 1. 1) (2. 1. 1. 4) +(2. 2) + (1. 1) + (1. 1. where the latter transform under the (2. 0) tensor (2. 1) (2. 1) (2. 1. 4. 1. 4. 2) +(2. 3. 1) +(1. 2) +(2. 2. 1. 1) + (3. 1. 1. 1. The ˜ ˜ second column lists the number of bosonic + fermionic states for each multiplet. 1) +(2. n) representations of U Sp(2N+ ) × U Sp(2N− ). 4. 1. 2. 1. 1) (3. 1) (3.

4) + (2. n) are assigned to the n-dimensional representation of SU (2) and the ˜ n-dimensional representation of U Sp(2N ). 1) + (3. This yields the 128 + 128 states of the (2. 1) + (3. 1) (5. 0) supersymmetry. we turn to the case of (N+ . 6) +(2. 14) +(3. 0) (3. Restricting ourselves to 32 supercharges and requiring the highest helicity to be a ﬁve-dimensional representation of one of the SU (2) factors. For a recent discussion of one of these theories. one can continue and classify representations for other values of (N+ . D = 6 supermultiplets without gravitons and gravitini with (N. see [10].16 Unity from Duality: Gravity. 0) (4. 14) (4. Finally. 1) + (3. so that the helicity group in six dimensions is restricted to SU (2) × 1 and the states are characterized as representations of U Sp(2N ). supersymmetry (1. 0) and (0. There is in fact a smaller supermultiplet. N− ). there are just four theories. . 48) Subsequently we discuss the case (N+ . 2) supergravity multiplet. summarized in Table 6. However. The theories based on these multiplets have only rigid supersymmetry. 2). These states transform according to representations of U Sp(4) × U Sp(4). 1) + (3. 1). a single (5. 42) fermions (4. They are all chiral. In principle. there are the chiral theories which contain neither graviton nor gravitino states. 0) (2. As is obvious from the construction that we have presented. so that we expect them to be inconsistent at the nonlinear level. The second column lists the number ˜ of bosonic + fermionic states for each multiplet. The states (n. 1) + (1. 1) tensor multiplet. Gauge Theory and Strings Table 6. The multiplets are identical to those that underly the ﬁve-dimensional N -extended supergravities. 1) + (1. 4) (4. 5) (5. N− ) = (2. this will inevitably lead to states transforming in higher-helicity representations. which we discard because it contains gravitini but no graviton states. 6) + (2. 27) +(1. The smallest supermultiplet is given by the tensor product of the smallest (2. 0) # 8+8 24 + 24 64 + 64 128 + 128 bosons (5. 1) +(3. 8) + (2. 2) (4. 0) tensor multiplet with the (0. Some of these multiplets will suﬀer from the fact that they have more than one graviton state. 1) highest-helicity state and at most 32 supercharges. N− ) = (2. 14) (5. 2) supermultiplets. Here a supergravity multiplet exists [21] and can be obtained from the product of the states of the (2.

Knowledge of the relevant groups HR is important and convenient in writing down the supermultiplets. Central charges and multiplet shortening are discussed in Section 2. it is the nonrelativistic variant of the R-symmetry group that was introduced previously. The shortened supermultiplets are known as BPS multiplets. (2.10) {Qα . The analysis of massive supermultiplets takes place in the restframe. For instance. It can also reveal certain relations between supermultiplets. in D = 4 spacetime dimensions. the spinor charges transform under the automorphism group of the supersymmetry algebra that commutes with the spin rotation group. even between supermultiplets living in spacetimes of diﬀerent dimension. However. Q† } = 2M δαβ . so that one obtains a Euclidean supersymmetry algebra. associated with spin. Table 7 shows the smallest massive representations for N ≤ 4 in D = 4 dimensions as an illustration.3 Massive supermultiplets 17 Generically massive supermultiplets are bigger than massless ones because the number of supercharges that generate the multiplet is not reduced. Obviously the nonrelativistic group can be bigger than its relativistic counterpart. while the nonrelativistic one is the group U Sp(2N ). This group will also be denoted by HR . as it is required to commute with a smaller group. This only happens for special values of these charges. unlike for massless supermultiplets where one-half of the supercharges vanishes. The states then organize themselves into representations of the rest-frame rotation group. and massive supermultiplets can be decomposed in terms of massless ones. in the presence of mass parameters the superalgebra may also contain central charges. More explicit derivations can be found in [4]. β Just as before. but sometimes there exists a relationship that is less trivial. In this section we assume that the central charges are absent. Obviously. which could give rise to a shortening of the representation in a way similar to what happens for the massless supermultiplets. In particular the N = 4 supermultiplet of Table 7 appears in many places and coincides with the massless N = 8 supermultiplet . the D = 4 massive multiplets shown in Table 7 coincide with the massless supermultiplets of chirally extended supersymmetry in D = 6 dimensions shown in Table 4. SO(D − 1). Clearly the states of given spin can be assigned to representations of the nonrelativistic group HR = U Sp(2N ) and decomposed in terms of irreducible representations of the relativistic R-symmetry group U (N ). de Wit: Supergravity 2.B.4. the relativistic R-symmetry group is equal to U (N ). which contains U (N ) as a subgroup according to 2N = N + N. supermultiplets living in higher dimensions can always be decomposed into supermultiplets living in lower dimensions. For instance. The supercharges transform as spinors under this group.

which in turn coincides with the spin rotation group for D = 4. Not surprisingly. Extending the number of supercharges beyond 16 will increase the minimal spin of a massive multiplet beyond spin-2. By employing such ﬁelds one may be able to deﬁne an oﬀ-shell representation. Unfortunately. one may conclude that conformal supergravity does not exist for more than 16 charges. This fact makes is much more diﬃcult to construct an extended variety of actions for these theories. according to which the ﬁeld degrees of freedom should comprise a massive supermultiplet (while the states that are described could be massless). The D = 5 and the chiral D = 6 massless supermultiplets are subject to the same helicity group SU (2). a fermion doublet (with eight degrees of freedom) and a triplet of auxiliary scalar ﬁelds (with three degrees of freedom). Since the number of eﬀective supercharges is equal in these cases and given by Qeﬀ = 16 (remember that only half of the charges play a role in building up massless supermultiplets). many theories do not possess (ﬁnite-dimensional) oﬀ-shell representations. where the transformations close upon (anti)commutation without the need for imposing ﬁeld equations. These states can be described by a ﬁeld theory in which the supercharges generate corresponding supersymmetry variations on the ﬁelds. In certain cases one can improve the situation by introducing extra ﬁelds which do not directly correspond to physical ﬁelds. Such representations are called on-shell representations. this multiplet coincides with the multiplet of the currents that couple to an N = 2 supersymmetric gauge theory. because the transformation rules are implicitly dependent on the action. In fact. Gauge Theory and Strings in D = 5 dimensions. There is an oﬀ-shell counting argument. For instance. The N = 4 multiplet in Table 7 corresponds to the gravitational supermultiplet of currents [22]. the oﬀ-shell description of the N = 2 vector multiplet in D = 4 dimensions can be formulated in terms of a gauge ﬁeld (with three degrees of freedom). Since higher-spin ﬁelds can usually not be coupled. So far we discussed supermultiplets consisting of states on which the supercharges act. for instance. These ﬁelds are known as auxiliary ﬁelds. The lack of closure has many consequences. Notorious examples are gauge theories and supergravity theories with 16 or more supercharges. These are the currents that couple to the ﬁelds of N = 4 conformal supergravity. Very often the transformations on the ﬁelds do not close according to the supersymmetry algebra unless one imposes the equations of motion for the ﬁelds. as the reader can easily verify. the multiplets must indeed be identical.18 Unity from Duality: Gravity. For that . which is shown in Tables 10 and 11. The reasons for this are clear. precisely in accord with the N = 2 entry in Table 7. when determining quantum corrections. also the relevant automorphism groups HR coincide. Here we also want to brieﬂy draw the attention to the relation between oﬀ-shell multiplets and massive representations.

These gauge transformations usually appear in the anticommutator of two supercharges. it is possible to have (part of) the automorphism group HR realized as a local invariance. some of the central charges may take ﬁnite values. local supersymmetry transformations and the gauge transformations associated with the antisymmetric gauge ﬁelds7 . Subsequently we then discuss some further implications of these results. not in perturbation theory). As we discuss in due course. . However. and may be regarded as central charges. In Section 2.B. dyons.5 we will present a table listing the various groups HR for spinors associated with certain Cliﬀord algebras C(p. However. the corresponding gauge ﬁelds are composite and do not give rise to physical states (at least. the theory does not contain charged ﬁelds. At the M-theory level. q). Without further knowledge about the kind of states that may emerge at the nonperturbative level. at the nonperturbative level. On such states. there may be solitonic or other states that carry charges. spin 1 1/2 0 2 3/2 1 1/2 0 N =1 1=1 2=1+¯ 1 N =3 1=1 6=3+¯ 3 14 = 8 + 3 + ¯ 3 14 = 6 + ¯ + 1 + ¯ 6 1 N =2 1=1 4=2+¯ 2 5=3+1+¯ 1 N =4 1=1 8=4+¯ 4 27 = 15 + 6 + ¯ 6 4 48 = 20 + 20 + 4 + ¯ 42 = 20 + 10 + 10 + 1 + ¯ 1 reason there can be no oﬀ-shell formulations for supergravity with more than 16 charges. de Wit: Supergravity 19 Table 7. In perturbation theory. The states are listed as U Sp(2N ) representations which are subsequently decomposed into representations of U (N ). 7 There may be additional gauge transformations that are of interest to us. Conformal supergravity will be discussed in Section 7. q) with corresponding rotation groups SO(p. or black holes. 2. these charges are associated with certain brane conﬁgurations. so these central charges simply vanish on physical states. An example are magnetic monopoles. Minimal D = 4 massive supermultiplets without central charges for N ≤ 4.4 Central charges and multiplet shortening The supersymmetry algebra of the maximal supergravities comprises general coordinate transformations.

1) +(5. The second row speciﬁes the number of independent components for each p-rank tensor charge. The total number of central charges is equal to 528 − D. 4) 1+2 [126] 3+¯ 3 [35] [15] 36 [10] 36 + 36 [3] we can generally classify the possible central charges. (2. 1) +(1. 1) +(1. because we have not listed the D independent momentum operators. 5) [6] 27 [5] 63 [4] 135 [3] p=2 1 [55] 1 [45] p=3 p=4 1 [210] 1 [120] 1 [84] 1+3 [56] 10 [35] (10. C is the charge-conjugation matrix and Zµ1 ···µp is the central charge. Qβ } ∝ (Γµ1 ···µp C)αβ Zµ1 ···µp . Decomposition of the central extension in the supersymmetry algebra with Q = 32 supercharge components in terms of p-rank Lorentz tensors. D 11 HR 1 p=0 p=1 10A 1 10B SO(2) 9 8 7 6 SO(2) U (2) U Sp(4) U Sp(4) ×U Sp(4) 1 [1] 1+2 [1] 3+¯ 3 [1] 10 [1] (4. 4) 5 4 3 U Sp(8) U (8) SO(16) [1] 1 + 27 [1] 28 + 28 [1] 120 [1] 1 [10] 2 [10] 2 [9] 3 [8] 5 [7] (1. 10) [10] p=5 1 [462] 1+1 [126] 1+2 [126] 1 [36] 1+¯ 1 [28] 1+5 [21] (4. This anticommutator carries at least two spinor indices and two indices associated with the group HR .11) p where Γµ1 ···µp is the antisymmetrized product of p gamma matrices.20 Unity from Duality: Gravity. Gauge Theory and Strings Table 8. which . by considering a decomposition of the anticommutator. Hence we may write {Qα .

The results of the table are in direct correspondence with the eleven-dimensional superalgebra with the most general central charges. In order to realize the supersymmetry algebra in a positive-deﬁnite Hilbert space. For BPS multiplets some of the supercharges must vanish on the states. For higher extended supersymmetry the diﬀerence in size of BPS supermultiplets and massive supermultiplets can be enormous in view of the fact that the number of states depend exponentionally on the number of nonvanishing central charges. For given spacetime dimension all possible central charges can be classiﬁed8 . for example [23–25] and references therein. For lower supersymmetry the multiplets can be 8 For related discussions see. Because we have 32 supercharge components.B. but here the fraction of the charges that vanishes is not necessarily equal to 1/2. depending on whether the (Γµ1 ···µp C)αβ is antisymmetric or symmetric in α. the bound may can take a complicated form. etcetera. The central charge must be symmetric or antisymmetric in these indices. β. Especially in higher dimensions. Hence one speaks of 1/2-BPS. so that the product with Zµ1 ···µp is always symmetric in the combined indices of the supercharges.12) The two central charges. Qβ } = −2iPM ΓM + ZMN ΓMN + ZMN P QR ΓMN P QR . This positivity bound is known as the Bogomol’nyi bound. the right-hand side of the anticommutator is subject to a positivity condition. the sum of the independent momentum operators and the central charges must be equal to (32 × 33)/2 = 528. ZMN and ZMN P QR can be associated with the winding numbers of two. The fact that the BPS supermultiplets have a completely diﬀerent ﬁeld content than the generic massive supermultiplets makes that they exhibit a remarkable stability under “adiabatic” deformations. Qualitatively. ¯ {Qα . This means that perturbative results based on BPS supermultiplets can often be extrapolated to a nonperturbative regime. This vanishing of some of the supercharges leads to a shortening of the multiplet. de Wit: Supergravity 21 transforms as an antisymmetric p-rank Lorentz tensor and depends on possible additional HR indices attached to the supercharges. to indicate which fraction of the supercharges vanishes on the states. 1/4-BPS supermultiplets. αβ αβ αβ (2. this phenomenon of multiplet shortening is the same as for massless supermultiplets.and ﬁve-branes. . For the maximal supergravities in spacetime dimensions 3 ≤ D ≤ 11 this classiﬁcation is given in Table 8. When the bound is saturated one speaks of BPS states. in the same way as half of the charges vanish for the massless supermultiplets. which generically implies that the mass of the multiplet is larger than or equal to the central charges. where we list all possible charges and their HR representation assignments.

. and ﬁve states with spin 0. ep+q . but nevertheless they are quite diﬀerent. which is BPS. Minkowski and (anti-)de Sitter spaces as well as the supersymmetry algebras based on these spinors. q) based on p + q generators. but with a completely diﬀerent spin content. These states are subject to a nonvanishing central charge which requires that the states are all doubly degenerate. 1 this square equals the identity. . The second supermultiplet. Therefore. the four states of two spin. the . In later sections we will discuss speciﬁc BPS supermultiplets as well as other mechanisms of multiplet shortening in anti-de Sitter space. . Without central charges. q). also known as the R-symmetry group. where r is deﬁned by r ≡ p − q mod 4. consider N = 2 massive vector supermultiplets in four spacetime dimensions. Hence this is the multiplet one has in the Higgs phase. . Subsequently we shall discuss some correspondences between the various spinors in diﬀerent dimensions. This multiplet carries no central charge.5 On spinors and the R-symmetry group HR In this section we return once more to the spinor representations and the corresponding automorphism group HR . q). 3 the square equals minus the identity. In view of the very large variety of BPS supermultiplets. while for r = 2. On the other 2 hand there is another massive vector supermultiplet. Gauge Theory and Strings comparable in size. and the values r (equal to 0. From this table we can gain certain insights into the properties of spinors living in Euclidean. . This means that p generators square to the identity and q to minus the identity.22 Unity from Duality: Gravity. for r = 0. We consider the Cliﬀord algebras C(p. For r = 0. . with a nondegenerate metric of signature (p. For instance. corresponding to the three states of spin-1. so that we have again 8 + 8 states.1 representations 2 and two states of spin-0. We list the real dimension of the irreducible Cliﬀord algebra representation. This realization is known as the Coulomb phase. Table 9 summarizes information for spinors up to (real) dimension 32 associated with the groups SO(p. denoted by e1 . we do not continue this general discussion of supermultiplets with central charges. denoted by dC . 3). Let us ﬁrst elucidate the information presented in the table. the eight states of four irreducible spin.1 representations. the ﬁrst multiplet decomposes into a massless vector multiplet and a hypermultiplet. where the hypermultiplet provides the scalar degree of freedom that allows the conversion of the massless to massive spin-1 states. 2. appears as a massive charged vector multiplet when breaking a non-Abelian supersymmetric gauge theory to an Abelian subgroup. . such a multiplet comprises 8 + 8 states. where we restrict q ≤ 2. . e2 . The value for r determines the square of the matrix built from forming the product e1 · e2 · · · ep+q of all the Cliﬀord algebra generators. which is BPS and comprises the three states of spin-1. When decomposing these multiplets into massless N = 2 supermultiplets.

This means that the irreducible representation commutes with the identity and none. consisting of the linear transformations that commute with the group SO(p. respectively. respectively. which leads to HR = SO(N ). The ﬁrst correspondence is between spinors of SO(p. When r = 0 the Cliﬀord algebra representation decomposes into two chiral spinors. leaving a positive-deﬁnite metric invariant. Table 9 reﬂects also the so-called periodicity theorem [26]. For r = 2. q) (or C(p. q) generated by products of even numbers of generators is not simple and breaks into two simple ideals. We also present the centralizer of the irreducible representations of the Cliﬀord algebra. one or three complex structures. or U Sp(2N ). Observe that for r = 2 we can also have chiral spinors. q) decomposes into two simple ideals. 0). the complex numbers (C). according to Schur’s lemma the centralizer of SO(p. which. de Wit: Supergravity 23 subalgebra C+ (p. or the quaternions (H). q) must form a division algebra for irreducible spinor representations. according to Schur’s lemma. q) spinor representations. q) and act on N irreducible spinors. 1) and SO(p − 1. q) commutes with the identity and none. q) diﬀers by a factor 16. a group HR should be assigned to each of the chiral sectors separately. must form a division algebra and is thus isomorphic to the real numbers (R). q) times the 16 × 16 real matrices. the group SO(p. respectively. which generate the corresponding division algebra. the dimensions of the corresponding spinors diﬀer by a factor two. 0) is the helicity group of massless spinor states in ﬂat Minkowski space of dimension D = p + 1. while their respective groups HR always coincide. Therefore. q +8)) and C(p. for any p > 1. the full Cliﬀord algebra C(p. The last column gives the compact group HR . Again. q) (or C(p. We note that the results of Table 9 are in accord with the results presented earlier in Tables 1 and 2. according to which there exists an isomorphism between the Cliﬀord algebras C(p+8. For r = 0. the dimension of the representations of C(p+8. one or three complex structures. Finally the table lists the branching of the Cliﬀord algebra representation into SO(p. while for r = 1. but they are complex so that their real dimension remains unaltered.3) thus share the same automorphism group. From a . but nevertheless in certain cases the Cliﬀord algebra representation can still decompose into two irreducible spinor representations. We now discuss and clarify a number of correspondences between spinors living in diﬀerent dimensions. Correspondingly. According to the table.B. simply coincide. From a physical perspective. this correspondence can be understood from the fact that SO(p − 1. In a ﬁeld-theoretic context the reduction of the spinor degrees of freedom is eﬀected by the massless Dirac equation and the automorphism groups HR that commute with the Lorentz transformations and the transverse helicity rotations.1) and (2. U (N ). The two algebras (2. 3 there are no chiral spinors. q+8)) and C(p.

2) C(4. 0) C(10. 1) C(7. 1) C(8. 0) C(8.q) 1 1+1 1+1 2 2 2+2 2+2 2+2 4 4 4 4+4 8 8 8 8+8 16 8+8 16 8+8 8+8 16 16 + 16 16 + 16 32 16 + 16 16 + 16 32 32 32 + 32 32 + 32 32 + 32 HR SO(N ) SO(N ) SO(N ) U (N ) SO(N ) U (N ) SO(N ) SO(N ) U Sp(2N ) U (N ) SO(N ) U Sp(2N ) U Sp(2N ) U (N ) U Sp(2N ) U Sp(2N ) U Sp(2N ) U (N ) U Sp(2N ) SO(N ) SO(N ) SO(N ) U Sp(2N ) U (N ) U Sp(2N ) SO(N ) SO(N ) U (N ) SO(N ) U (N ) SO(N ) SO(N ) . 0) C(6. 2) C(2. q) spinors of maximal real dimension 32 and their Rsymmetry group. 0) C(4. 0) C(2. 2) C(9. 1) C(1. 1) C(5. 0) C(6. 2) C(7. 2) C(3.24 Unity from Duality: Gravity. 1) C(3. 2) C(10. Gauge Theory and Strings Table 9. 1) C(0. 1) C(4. q) C(1. 0) C(5. 0) C(0. 1) C(9. 2) C(6. 1) C(2. 1) C(10. 2) C(1. 2) C(5. Representations of the Cliﬀord algebras C(p. q) with q ≤ 2 and their centralizers. dC 1 2 2 2 2 4 4 4 4 4 4 8 8 8 8 16 16 16 16 16 16 16 32 32 32 32 32 32 32 64 64 64 C(p. as well as r = p − q mod 4. and the SO(p. We also list the dimensions of the Cliﬀord algebra and spinor representation. 0) C(3. 1) C(7. 2) C(8. 0) C(9. 2) r 1 3 0 2 1 2 3 0 3 2 1 0 3 2 1 0 3 2 1 3 0 1 0 2 3 3 0 2 1 2 3 0 centralizer R C R R R H C R H C R H H C H H C H H C R R H C H C R R R H C R dSO(p.

an underlying isomorphism for the corresponding Cliﬀord algebras is lacking. 1) and SO(p.e. 5.13) where we note that C(1. a single Minkowski spinor can not be elevated to anti-de Sitter space. The third correspondence relates spinors of SO(p. 1) the R-symmetry group is larger and equal to U Sp(2N ). 4. Therefore theories formulated in ﬂat Minkowski spacetime of dimension D = 3. In that case the dimension of the automorphism group HR can increase. q − 1) ⊗ C(1. the relativistic spinor decomposes into two nonrelativistic spinors. 4. while for D = 3. Again the situation depends sensitively on the value for p. q) ∼ C(p − 1. which are proportional to Γ[a Γb] . which transform in the restframe under p-dimensional spin rotations. In the remaining cases. as it is based on the isomorphism (2. In the remaining dimensions. as it does for p = 3 and 10. Inspired by the ﬁrst correspondence one may investigate a second one between spinors of SO(p. 2) spinors allow the R-symmetry group U (N ). 6. The results of the table should therefore be applied with care. 1) with p > 1. p = 2. equivalently. so that one can always extend the generators of SO(p. Because the number of irreducible spinors is then doubled. 8 (always modulo 8). de Wit: Supergravity 25 mathematical viewpoint. 9. 10..3. In a number of cases (i. p = 4. = (2. and one must at least start from an even number of ﬂat Minkowski spinors (so that N is even). In a number of cases the relativistic spinor transforms irreducibly under the nonrelativistic rotation group. 2) . The correspondence relates spinors of SO(p. 6) the spinor dimension is the same for both groups. the Rsymmetry group is not necessarily the same. but one should consult the table for speciﬁc cases. Physically this correspondence is relevant when considering relativistic massive spinors in ﬂat Minkowski spacetime of dimension D = p + 1. 1) . This can be understood from the fact that the Cliﬀord algebra representations are irreducible with respect to SO(p. to those of SO(p. while for SO(4.13). indeed. 6 the R-symmetry remains the same. 8. this correspondence is related to the isomorphism C(p. For p = 3 (or. 6 can in principle be elevated to anti-de Sitter space. 0) with p > 1. 4. have already been discussed in Section 2. 7. 7. For p = 4 the SO(4. The fourth correspondence is again more systematic. D = 2. 2) by including the gamma matrices Γa . 3. 1) is isomorphic with the real 2 × 2 matrices. 1). this correspondence is less systematic and. 1). 2) and SO(p. D = 4) the implications of the fact that the nonrelativistic automorphism group U Sp(2N ) is bigger than the relativistic one. it is hard to make general statements about the fate of the R-symmetry when moving to anti-de Sitter space and one has to consult Table 9. the nonrelativistic automorphism group has a tendency to increase. For D = 5 the R-symmetry reduces to U (N ). As the table shows. However. but not for p = 5 and 9.B. For these cases.

4)p .26 Unity from Duality: Gravity. 2) can be regarded as the group of conformal symmetries in a Minkowski space of p dimensions. or as the isometry group of an anti-de Sitter space of dimension p + 1. 2 2 where p indicates the three-momentum. Gauge Theory and Strings and of SO(p − 1. In the anti-de Sitter context. Multiplying this multiplet with a similar one. the spinor charge is irreducible but has simply twice as many components. We close this section by exhibiting a chain of relationships between various supermultiplets. (2. the U (1) subgroup of U (4) coincides with the helicity group and plays no independent role here. of which only 8 are realized on the massless supermultiplet. |p| the energy. We return to the superconformal invariance and related aspects in Section 7. 1)p + (0. The anticommutator of two S-supersymmetry charges yields the conformal boosts.14) were only subject to the helicity group and 8 supersymmetries. Observe that SO(p. but now with opposite three-momentum −p. and the entries in the parentheses denote the helicity and the SU (4) representation of the states. It implies that the extension of the Poincar´ superalgebra in e D = p spacetime dimensions to a superconformal algebra requires a doubling of the number of supercharges. While the states of the original multiplet (2. which plays a more basic role in the superconformal algebra as its generators appear in the anticommutator of a Q-supersymmetry and an S-supersymmetry charge. yields a multiplet with zero momentum and with mass M = 2|p|. The resulting supermultiplet consists of 128 + 128 degrees of freedom. 6)p + ( 1 . For all p > 1 the spinor dimension diﬀers by one-half while the R-symmetry group remains the same. . We start with an N = 4 supersymmetric gauge theory in D = 4 spacetime dimensions. It is illuminating to exploit some of the previous correspondences and the relations between various supersymmetry representations in the context of the so-called adS/CFT correspondence [28]. whose massless states are characterized as representations of the SO(2) helicity group and the R-symmetry group SU (4)9 . 4)p + (− 1 . 9 Because this multiplet is CPT self-conjugate. so that they can be characterized by their spin. Hence we have a ﬁeld theory with Q = 16 supersymmetries.and S-supersymmetry. 1). This supermultiplet decomposes as follows. This correspondence extends this statement to the level of spinors. Both set of charges transform under the R-symmetry group of the Poincar´ e algebra. This feature is well known [27] and the two supersymmetries are called Q.14) (±1. As it turns out the helicity states can now be assembled into states that transform under the 3-dimensional rotation group. the composite multiplet is now a full supermultiplet subject to 16 supersymmetries and the rotation (rather than the helicity) group.

divergence-free) and traceless. The N = 4 conformal supergravity theory couples consistently to the N = 4 supersymmetric gauge theory. it forms the basis for a proper oﬀ-shell theory of N = 4 conformal supergravity [22]. The presence of the traceless and conserved energy-momentum tensor and supersymmetry currents. Hence this is the same multiplet that describes D = 5 maximal supergravity. the states are annihilated by half the supercharges and are still classiﬁed according to SO(3). de Wit: Supergravity 27 Indeed. It is possible to cast the above product of states into a product of ﬁelds of the 4-dimensional gauge theory. Because neither the currents nor the conformal supergravity ﬁelds are subject to any ﬁeld equations (unlike the supersymmetric gauge multiplet from which we started. The oﬀ-shell N = 4 conformal supergravity multiplet in 4 dimensions can also be interpreted as an on-shell massless supermultiplet in 5 dimensions with 32 supersymmetries. and 6 selfdual antisymmetric tensors. Furthermore there are 4 chiral and 4 antichiral vector-spinor operators. The spin-1 operators decompose into 15 conserved vectors. anti-de Sitter space leads to “remarkable representations”. which are conserved and traceless (with respect to a contraction with gamma matrices) such that each of them correspond precisely to the 4 components appropriate for spin. 2 1 Finally there are 20 and 4 chiral and antichiral spin. is a consequence of the superconformal invariance of the underlying 4-dimensional gauge theory. which couples to the ﬁelds of conformal supergravity. the R-symmetry group coincides with the U Sp(8) R-symmetry of the relativistic 5-dimensional supersymmetry algebra. These are the singletons. In this form the relevant R-symmetry group is extended to U Sp(8). which is conserved (i. As we will discuss in Section 6. which do . which constitutes only an on-shell supermultiplet). According to Table 9.2 operators. 5 for a discussion of this) in which case it can possess an anti-de Sitter ground state.B. associated with the currents of SU (4). This theory has a nonlinearly realized E6(6) invariance whose linearly realized subgroup (which is relevant for the spectrum) equals U Sp(8). so that it has precisely the 5 independent components appropriate for spin-2. The spin-0 operators are scalar composite operators.e. and of the SU (4) conserved currents. a fully supersymmetric ground state leads to a U (4) R-symmetry group.3 . inspection shows that this composite multiplet is precisely the N = 4 massive multiplet shown in Table 7. One then ﬁnds that the spin-2 operators correspond to the energy-momentum tensor. These are the supersymmetry currents. Section 7 will further explain the general setting of superconformal theories that is relevant in this context. This is precisely the supermultiplet of currents [22]. which now acts as the helicity group. The latter theory can be gauged (we refer to Sect. Because of the masslessness.

their symmetries. Therefore the gauge group must be equal to SO(6) ∼ SU (4)/Z2 . namely a vielbein ﬁeld eµ and a ab spin-connection ﬁeld ωµ . Because the 5dimensional anti-de Sitter superalgebra coincides with the 4-dimensional superconformal algebra. We exhibit the initial steps in the construction of a supergravity theory.1 Simple supergravity The ﬁrst steps in the construction of any supergravity theory are usually based on the observation that local supersymmetry implies the invariance under general coordinate transformation. At this stage we thus have to content ourselves with the existence of this remarkable chain of correspondences. with and without a cosmological term. and dimensional compactiﬁcations on tori. Therefore one must introduce the a ﬁelds needed to describe general relativity.. is purely based on symmetries. This set-up requires the gauge group of 5-dimensional supergravity to be chosen such as to preserve the relevant automorphism group. the circle closes. For an introduction . Then we concentrate on maximal supergravity theories in various dimensions. Thus. while the spin-connection ﬁeld is associated with (local) Lorentz transformations of these frames. Hence it does not come as a surprise that these singleton representations coincide with the supermultiplet of 4dimensional N = 4 gauge theory. Therefore these theories are necessarily theories of gravity. .. . linking the various supermultiplets in diﬀerent dimensions by a series of arguments. b. The vielbein ﬁeld is nonsingular and its inverse µ is denoted by ea . . the 4-dimensional boundary theory must be consistent with superconformal invariance. Many aspects of these correspondences will reappear in later sections. this gauging allows for an anti-de Sitter max= imally supersymmetric ground state [29].28 Unity from Duality: Gravity. and the tangent space indices. . . Indeed. Because of the underlying supersymmetry algebra. We stress that the above excursion. ν. The vielbein deﬁnes a local set of tangent frames of the spacetime manifold. At the end we brieﬂy discuss some of the nonmaximal theories 3. . both run from 0 to D−1. 3 Supergravity In this section we discuss ﬁeld theories that are invariant under local supersymmetry. a. The world indices. Gauge Theory and Strings not have a smooth Poincar´ limit because they are associated with possible e degrees of freedom living on a 4-dimensional boundary. the invariance under local supersymmetry implies the invariance under spacetime diﬀeomorphisms. µ. It does not capture the dynamical aspects of the adS/CFT correspondence and has no bearing on the nature of the gauge group in 4 dimensions.

ab D[µ (ω)Rνρ] (ω) = 0 . but this restriction is not essential10 . de Wit: Supergravity 29 to the vielbein formalism we refer to [30].5) 10 For deﬁniteness we consider a generic supergravity theory with one Majorana gravitino with an antisymmetric charge-conjugation matrix C and gamma matrices Γa satisa fying CΓa C −1 = −ΓT .4) a ab It is suggestive to regard eµ and ωµ as the gauge ﬁelds of the Poincar´ e group. Hence any supergravity Lagrangian is expected to contain the Einstein-Hilbert Lagrangian of general relativity and the Rarita-Schwinger Lagrangian for the gravitino ﬁeld. ¯ κ2 L = − 1 e R(ω) − 1 eψµ Γµνρ Dν (ω)ψρ + · · · . so that P and M denote the translation and the Lorentz generators of the Poincar´ algebra.B. denoted by ψµ . Γb ] are the generators of the Lorentz 2 4 transformations in spinor space. Γ[µν] . ab ab ab ac b ac b Rµν (ω) = ∂µ ων − ∂ν ωµ + ωµ ων c − ων ωµ c . (3. Here we will just use the notation R(ω) and deﬁne its contractions with the inverse vielbeine (related to the Ricci tensor and Ricci scalar) by a ab Rµ (e.2) and ωµ ab is the spin-connection ﬁeld whose deﬁnition will be discussed in a sequel. We will e use this notation in later sections when discussing the anti-de Sitter and the conformal algebras. κ2 is related to Newton’s constant and a e = det(eµ ). 4 (3. namely the curvature ab a associated with the spin connection Rµν (ω) and the torsion tensor Rµν (P ). as it should [30]. Furthermore Γµ = eµ Γa . The matrices 1 Γab = 1 [Γa . a a a Rµν (P ) = Dµ (ω)eν − Dν (ω)eµ . We note the existence of two covariant tensors. Furthermore one needs one or several gravitino ﬁelds. (3. For simplicity we only consider a single Majorana gravitino ﬁeld. 2 2 where the covariant derivative on a spinor ψ reads ab Dµ (ω)ψ = ∂µ − 1 ωµ Γab ψ . which carry both a world index and a spinor index and which act as the gauge ﬁelds associated with local supersymmetry. ω) = ea ebν Rµν (ω) . a . (3. which carries this name because it is proportional to the antisymmetric part ρ of the aﬃne connection. Observe that the spinor covariant derivative on ψµ contains no aﬃne connection. ω) = ebν Rµν (ω) .1) (3. In that context R(ω) is written as R(M ).3) We note that these tensors satisfy the Bianchi identities. a ab D[µ (ω)Rνρ] (P ) + R[µν (ω) eρ] b = 0 . upon using the vielbein postulate. µ ab R(e.

which ensures the validity of the vielbein postulate. With the zero-torsion value (3. For pure gravity the ﬁrst.and the second-order formulism lead to the same result.30 Unity from Duality: Gravity. The supersymmetry transformations contain the terms. whose existence can be inferred from the knowledge of the possible underlying (massless) supermultiplets of states. (3. 2 (3. µ ν c c Ωab c = ea eb (∂µ eν − ∂µ eν ) .6) can be solved algebraically and leads to. which is then solved in terms of its ﬁeld equations. (3.6) Such a constraint is called “conventional” because it expresses one ﬁeld in terms of other ﬁelds in an algebraic fashion. Extending this Lagrangian to a fully supersymmetric one is not always possible. for instance. conventional supergravity no longer exists.7) where the Ωab c are the objects of anholonomity. The constraint (3.11) . When the spacetime dimension exceeds eleven. or it can be ﬁxed from the beginning (second-order formalism). up to terms that are cubic in the gravitino ﬁeld. ω) − 1 eψµ Γµνρ Dν (ω)ψρ 2 2 ¯ + 1 g(D − 2)e ψµ Γµν ψν + 1 g 2 (D − 1)(D − 2) e + · · · (3. as we shall discuss in the next section. (3.9) where the gravitino variation is the extension to curved spacetime of the spinor gauge invariance of a Rarita-Schwinger ﬁeld. δeµ a = 1 ¯ Γa ψµ . The action corresponding to the above Lagrangian is locally supersymmetric up to terms cubic in the gravitino ﬁeld.10) 4 2 As it turns out the corresponding action is still locally supersymmetric. Gauge Theory and Strings The spin connection can be treated as an independent ﬁeld (ﬁrst-order formalism). Usually it requires additional ﬁelds of lower spin. (3. δeµ a = 1 ¯ Γa ψµ . provided that we introduce an extra term to the transformation rules. Let us now include a cosmological term into the above Lagrangian as well as a suitably chosen masslike term for the gravitino ﬁeld.7) for the spin connection. 2 δψµ = Dµ (ω) + 1 gΓµ 2 . L ¯ = − 1 e R(e. by imposing the constraint. 2 δψµ = Dµ (ω) . the aﬃne connection ab becomes equal to the Christoﬀel symbols and Rµνρ σ = Rµν (ω) eρ a ebσ coincides with the standard Riemann tensor. ab c ωµ (e) = 1 eµ (Ωab c − Ωb c a − Ωc ab ) . a Rµν (P ) = 0 .8) From the spin connection one deﬁnes the aﬃne connection by Γµνρ = ρ ea Dµ (ω)eνa .

where ωµ and eµ act as gauge ﬁelds11 . i. With the conventions that we have adopted this implies that g is real. Assuming that the theory has an anti-de Sitter or de Sitter ground state.10) reads (suppressing the gravitino ﬁeld). R = g 2 D(D − 1) . 2 . simple supergravity in D = 10 dimensions cannot possibly have (anti-)de Sitter ground states. unless one accepts “ghosts”: ﬁelds whose kinetic terms are of the wrong sign. but it must be of deﬁnite sign (at least. but there are also explicit studies ruling out supersymmetric cosmological terms in 11 dimensions [35]. Observe that the variation for ψµ may be regarded as a generalized ab a covariant derivative. For instance. Therefore.e. if the ground state is to preserve supersymmetry).B. Here the argument may be invoked that no relevant supersymmetric extension of the anti-de Sitter algebra exists beyond D = 7 dimensions [3]. Rµν − 1 gµν R + 1 g 2 (D − 1)(D − 2) gµν = 0 . 1 ¯ − 2 eψµ Γµνρ (Dν (ω) + 1 gΓν )ψρ . as it implies that the cosmological term is of deﬁnite sign. as one does not always have a single Majorana spinor with the speciﬁed charge conjugation properties. de Wit: Supergravity 31 The Lagrangian (3.. 34] and references therein. Nevertheless the conclusion that the cosmological term must have this particular sign remains. 2 2 which implies. Rµν = g 2 (D − 1) gµν . The Einstein equation corresponding to (3.10) is consistent with that interpretation as it can be generated from the Rarita-Schwinger Lagrangian by the same change of the covariant derivative. Consistency requires that gΓµ satisﬁes the same Majorana constraint as ψµ and . Hence supersymmetry does not a priori forbid a cosmological term. see [33.10) was ﬁrst written down in [31] in four space-time dimensions and the correct interpretation of the masslike term was given in [32]. one may verify whether the Minkowski spinors have the right dimension to enable them to live in these spaces. Such a counting argument does not exclude anti-de Sitter ground states in D = 11 spacetime dimensions.12) 11 The masslike term in (3. because D = 11 Lorentz spinors can exist in anti-de Sitter space. We should point out that there are situations where a cosmological term is not consistent with supersymmetry. The reality of g has important consequences. This example does not cover all cases. a MajoranaWeyl spinor in D = 10 spacetime dimensions has only half the number of components as compared to a spinor in (anti-)de Sitter space of the same dimension.13) (3. For an early discussion. (3.

where we discuss the (super)multiplet structure in anti-de Sitter space. (3.15) holds for any spinor .2 Maximal supersymmetry and supergravity In Section 2 we restricted ourselves to supermultiplets based on Q ≤ 32 supercharge components. The need for certain extra ﬁelds can be readily deduced from the underlying massless supermultiplets.14) This solution leaves all supersymmetries intact just as ﬂat Minkowski space does. (3. The maximally symmetric solution of this equation is an anti-de Sitter space. For instance.14) so that the spinor must live in anti-de Sitter space.15) Spinors satisfying this equation are called Killing spinors. in this section.16) from which one derives that the Riemann tensor satisﬁes (3. We will return to this issue later in Section 6. Gauge Theory and Strings Hence we are dealing with a D-dimensional Einstein space.13). implies (3. in D = 11 dimensions an additional antisymmetric gauge ﬁeld is necessary. Antisymmetrizing this 2 2 expression in µ and ν then yields the (algebraic) integrability condition − 1 Rµνab Γab + 1 g 2 Γµν 4 2 Multiplication with Γν yields Rµν − g 2 (D − 1) gµν Γν = 0 . which were extensively discussed in the previous section. Since (3. One can verify this directly by considering the supersymmetry variation of the gravitino ﬁeld and by requiring that it vanishes in the bosonic background. This happens for spinors (x) satisfying Dµ (ω) + 1 gΓµ 2 = 0. We stress once more that. 3. (3. so that (3.15) is a ﬁrst-order diﬀerential equation. Therefore supersymmetry requires an Einstein space. Requiring full supersymmetry. To construct the full theory usually requires more ﬁelds and important restrictions arise on the dimensionality of spacetime. while minimal supergravity in D = 4 dimensions does not require additional ﬁelds. we have restricted ourselves to the graviton-gravitino sector. (3. To see this one notes that also (Dµ (ω) + 1 gΓµ )(Dν (ω) + 1 gΓν ) must vanish. Hence we have seen that supersymmetry can be realized in anti-de Sitter space. From the general analysis it is clear that increasing the number of supercharges leads to higher and higher helicity . one expects that it can be solved provided some integrability condition is satisﬁed.17) = 0. whose Riemann curvature equals Rµνab = 2g 2 eµ[a eν b] .32 Unity from Duality: Gravity.

B. Those are described by gauge ﬁelds that are symmetric Lorentz tensors. upon dimensional reduction. Beyond Q = 32 one is dealing with states of helicity λ > 2. An exception is the graviton ﬁeld. so that the theory should contain Rarita-Schwinger ﬁelds. In 2 view of the supersymmetry algebra an interacting supersymmetric theory of this type should contain gravity. which can interact with itself as well as with low-spin matter. (ii) interactions that are inversely proportional to the cosmological constant. Hence the conclusion is that there is a restriction on the number Q of independent supersymmetries. There have been many eﬀorts to circumvent this bound of Q = 32 supersymmetries. neither to themselves nor to other ﬁelds. However. see e. because otherwise. of a massless supermultiplet in D = 4 spacetime dimensions is larger than or equal to 1 16 Q. the theory does not support more than 32 supercharges. see. the maximal helicity. However. [39]): (i) an inﬁnite tower of higher-spin gauge ﬁelds. for instance [36]). When Q > 16 we have |λmax | ≥ 3 . when coupling matter to this theory in the form of scalars and spinors. 20.e. By consistent. but in higher dimensional spacetimes one expects to arrive at the same conclusions. these theories would give rise to theories that are consistent in D = 4. Indeed explicit theories have been constructed which demonstrate this. This is the reason why we avoided (i. Therefore.g. one can write down supergravity theories based on a graviton ﬁeld and an arbitrary number of gravitino ﬁelds. it turns out that symmetric gauge ﬁelds cannot consistently couple. but not with other ﬁelds of the same spin [37]. Most of the search for interacting higher-spin ﬁelds was performed in 4 spacetime dimensions [38]. Symmetric tensor gauge ﬁelds for arbitrary helicity states can be constructed (in D = 4 dimensions. so that in this case we must include λ = 2 states for the graviton. conventional supergravity theories are not of this kind. 24 and 32 [8]. Beyond Q = 16 there are four unique theories with Q = 18. The fact that an inﬁnite number of ﬁelds can cure certain inconsistencies is by itself not new. where graviton and gravitini ﬁelds do not describe dynamic degrees of freedom. so that theories for these multiplets must include vector gauge ﬁelds. |λmax |. (iii) extensions of the super-Poincar´ or the super-de Sitter algebra with additional fermionic e and bosonic charges. Hence. we mean that the respective gauge invariances of the higher-spin ﬁelds (or appropriate deformations thereof) cannot be preserved at the interacting level. as for Q > 32 no interacting ﬁeld theories seem to exist. de Wit: Supergravity 33 representations. when Q > 8 we have |λmax | ≥ 1. For instance. While a massive spin-2 ﬁeld cannot be coupled to gravity. . It seems clear that one needs a combination of the following ingredients in order to do this (for a review. which are topological. There is also direct evidence in D = 3. However. in Table 5) to list supermultiplets with states transforming in higher-helicity representations.

where the extra dimensions can not uniformly decompactify so that the no-go theorem is avoided. Observe. D 11 10A 10B 9 8 7 6 5 4 3 HR 1 1 SO(2) SO(2) U (2) U Sp(4) U Sp(4) ×U Sp(4) U Sp(8) U (8) SO(16) graviton 1 1 1 1 1 1 1 1 1 1 p = −1 0 1 2 2+1 5+1+¯ 1 14 (5. The p = 3 gauge ﬁeld in D = 10B has a self-dual ﬁeld strength. there are scenarios based on spacetime dimensions higher than D = 11. These theories have Q = 32 supersymmetries and we restrict our discussion to 3 ≤ D ≤ 11. respectively) are extended to U (1) and SU (8) representations through duality transformations on the ﬁeld strengths. Therefore. 1) +(1. For p > 0 the ﬁelds can be assigned to representations of a bigger group than HR . These transformations cannot be represented on the vector potentials. The bosonic ﬁelds always comprise the metric tensor for the graviton and a number of (p + 1)-rank antisymmetric gauge ﬁelds. The fact that no uniform decompactiﬁcation is possible is closely related to the T -duality between winding and momentum states that one knows from string theory. 4) 27 [28] p=1 0 1 2 2 3 5 (5. The highest dimension D = 11 is motivated by the fact that spinors have more that 32 components in ﬂat Minkowski space for spacetime dimensions D > 11. In this section we review the maximal supergravities in various dimensions. but it turns out that the interactions often prefer the rank of the gauge ﬁeld to be as small as possible. as can be seen in Kaluza-Klein theory. the graviton does not describe propagating degrees of freedom. This will be discussed in due course. it is a priori unclear whether to choose a (p+1)-rank gauge ﬁeld or its dual (D − 3 − p)-rank partner. The representations [1] and [28] (in D = 8. . 5) 42 35 + 35 128 p=0 0 1 0 2+1 3+¯ 3 10 (4. In D = 3 dimensions. 4. that this argument assumes D-dimensional Lorentz invariance. however.34 Unity from Duality: Gravity.41]. For the antisymmetric gauge ﬁelds. Gauge Theory and Strings Table 10. 5) p=2 1 1 0 1 [1] p=3 0 0 1∗ the coupling of an inﬁnite number of them can be consistent. Bosonic ﬁeld content for maximal supergravities. As was stressed in [40.

the maximal supergravity theories have symmetry groups that are much larger than HR . Its Lagrangian can be written as . In D = 3 dimensions the gravitino does not describe propagating degrees of freedom. Obviously. 3. Actually. Note that the simplest versions of supergravity (which depend on no other coupling constant than Newton’s constant) are manifestly invariant under HR . we restrict ourselves to p ≤ 3. 6. 0) supergravity in 6 dimensions. always consisting of gravitini and simple spinors. because the only symmetric gauge ﬁeld is the one describing the graviton. In Table 11 we also present the fermionic ﬁelds. as we will explain in a sequel. 8 we include both chiral and antichiral spinor components.3 D = 11 supergravity A Supergravity in 11 spacetime dimensions is based on an “elfbein” ﬁeld EM . D 11 10A 10B 9 8 7 6 5 4 3 HR 1 1 SO(2) SO(2) U (2) U Sp(4) U Sp(4)×U Sp(4) U Sp(8) U (8) SO(16) gravitini 1 1+1 2 2 2+¯ 2 4 (4. 4) 8 8+¯ 8 16 spinors 0 1+1 2 2+2 2+¯+4+¯ 2 4 16 (4. 1) + (1. which transform in conjugate representations of HR . as in D = 11 dimensions. 5) + (5. 7 the fermion ﬁelds are counted as symplectic Majorana spinors. With chiral (2. For D = 5. All these ﬁelds are classiﬁed as representations of the R-symmetry group HR . This table presents all the ﬁeld conﬁgurations for maximal supergravity in various dimensions. Fermionic ﬁeld content for maximal supergravities. de Wit: Supergravity 35 Table 11. the problematic higher-spin ﬁelds are avoided. p = 3 and p = 4 are each other’s dual conjugates.B. a Majorana gravitino ﬁeld ΨM and a 3-rank antisymmetric gauge ﬁeld CMN P . 4) 48 56 + 56 128 in Table 10. For D = 4. it is the only Q ≥ 16 supergravity theory without a scalar ﬁeld.

(3. Note the presence of a Chern-Simons-like term F ∧ F ∧ C in the Lagrangian. We have the following bosonic ﬁeld equations and Bianchi identities.22) N] 4 The left-hand side is the supercovariant torsion tensor. in which the spin connection is deﬁned as a dependent ﬁeld determined by its (algebraic) equation of motion. The Lagrangian is derived in the context of the so-called “1. Gauge Theory and Strings follows [11]. √ = − 1 2 ¯ Γ[MN ΨP ] .18) by replacing the spinˆ connection ﬁeld Ω by (Ω + Ω)/2 in the covariant derivative of the gravˆ itino kinetic term and by replacing FMN P Q in the last line by (FMN P Q + FMN P Q )/2. Ω) − 1 E ΨM ΓMN P DN (Ω)ΨP − 48 E(FMN P Q )2 2 2 κ2 11 √ 1 − 3456 2 εMN P QRST UV W X FMN P Q FRST U CV W X (3.18) are supercovariant.23) .5-order” formalism.19) ˆ FN P QR . FP QRS F P QRS − 1 FMP QR FNP QR . 2 (3. A where the ellipses denote terms of order Ψ4 . whereas its supersymmetry variation in the action is treated as if it were an independent ﬁeld [42]. ¯ ˆ D[M (Ω) E A − 1 ΨM ΓA ΨN = 0 . 6 0.21) The supercovariant spin connection is the solution of the following equation. RMN ∂M E F MN P Q ∂[M FN P QR] = = = 1 72 gMN 1 1152 √ N P QRST UV W XY 2ε FRST U FV W XY . L11 = 1 1 ¯ − 1 E R(E. These substitutions ensure that the ﬁeld equations corresponding to (3. 8 √ N 1 ˆ = DM (Ω) + 288 2 ΓMN P QR − 8 δM ΓP QR 1 2 (3. We also wish to point out that the quartic-Ψ terms can be included into the Lagrangian (3. Here the covariant derivative is covariant with respect to local Lorentz transformation DM (Ω) = ∂M − 1 ΩM AB ΓAB .18) √ 1 ¯ ¯ − 192 2E ΨR ΓMN P QRS ΨS + 12 ΨM ΓN P ΨQ FMN P Q + · · · . The supersymmetry transformations are equal to A δEM = δCMN P δΨM ¯ ΓA Ψ M . (3.20) 4 ˆ and FMN P Q is the supercovariant ﬁeld strength √ ˆ ¯ FMN P Q = 24 ∂[M CN P Q] + 3 2 Ψ[M ΓN P ΨQ] . E = det EM and ΩMAB denotes the spin connection.36 Unity from Duality: Gravity. so that the action is only invariant under tensor gauge transformations up to surface terms. (3.

is undetermined and depends on ﬁxing some length scale.and a 5-dimensional spatial volume. the integral H may be associated with electric ﬂux and the integral F with magnetic ﬂux.B. the constant 1/κ2 in front of the Lagrangian (3. κ2 → e(2−D)α κ2 .26) Under this rescaling the Lagrangian changes according to L11 → e−9α L11 . These are just the charges that can appear as central charges in the supersymmetry algebra (2. respectively. √ 1 E εMN P QRST UV W X F UV W X − 1 2 F[MN P Q CRST ] .and 5-rank Lorentz tensors. κ2 → e−9α κ2 .12). To see this consider a continuous rescaling of the ﬁelds. Concentrating on the Einstein-Hilbert action in D spacetime dimensions. 2 7! (3. An alternative form of the second equation is [43] ∂[M HN P QRST U] = 0 .24) receive contributions from charges that would give rise to source terms on the right-hand side of the equations. (3. In analogy with the Maxwell theory. CMN P → e−3α CMN P . the corresponding scaling property is D D gµν → e−2α gµν . Finally. which 11 carries dimension [length]−9 ∼ [mass]9 . and the corresponding charges are 2.28) (3.24) ΨM → e−α/2 ΨM . The same situation is present in many other supergravity theories. de Wit: Supergravity 37 which no longer depend explicitly on the antisymmetric gauge ﬁeld.18). LD → e(2−D)α LD . . 11 11 (3. respectively. HMN P QRST = A A EM → e−α EM .25) One could imagine that the third equation of (3. The spatial volumes are orthogonal to a p = 2 and a p = 5 brane conﬁguration. These charges are associated with the “ﬂux”-integral of HMN P QRST and FMN P Q over the boundary of an 8.27) This simply means that the Lagrangian depends on only one dimensional coupling constant. Solutions of 11-dimensional supergravity that contribute to these charges were considered in [44–46]. provided the supersymmetry parameter is changed accordingly. (3. where HMN P QRST is the dual ﬁeld strength. namely κ11 . This change can then be absorbed into a redeﬁnition of κ11 12 .29) 12 Note that the rescalings also leave the supersymmetry transformation rules unchanged.23) and (3. D D (3.

. D − 1. Gauge Theory and Strings Of course. A subset of the gauge symmetries associated with the compactiﬁed dimensions survive as internal symmetries. For compactiﬁcations on less trivial spaces than the hypertorus. 1. which is an obvious requirement for having consistent truncations to the 13 Throughout these lectures we enumerate spacetime coordinates by 0. In this reduction some of the spatial dimensions are compactiﬁed on a hypertorus and one retains only the ﬁelds that do not depend on the torus coordinates. The massless modes form the basis of the lower-dimensional supergravity theory.38 Unity from Duality: Gravity. This results in a spectrum of massless modes and an inﬁnite tower of massive modes with masses inversely proportional to the circle length L. 3.4 Dimensional reduction and hidden symmetries The maximal supergravities in various dimensions are related by dimensional reduction. Because a toroidal background does not break supersymmetry. For instance. 11-dimensional supergravity can be compactiﬁed to a 4-dimensional maximally symmetric spacetime in only two ways such that all supersymmetries remain unaﬀected [47]. it is important to decompose the higher-dimensional ﬁelds in such a way that they transform covariantly under the lower-dimensional gauge symmetries and under diffeomorphisms of the lower-dimensional spacetime. . This ensures that various complicated mixtures of massless modes with the tower of massive modes will be avoided. In the latter case the resulting 4-dimensional supergravity theory acquires a cosmological term. the other one the compactiﬁcation on a sphere S 7 . In the formulation of the compactiﬁed theory. Fully supersymmetric compactiﬁcations are rare. This corresponds to the theory one obtains when the size of the torus is shrunk to zero. this is usually not the case and the number of independent supersymmetries will be reduced. The aim of the present discussion here is to elucidate a number of features related to these symmetries. the resulting supergravity has the same number of supersymmetries as the original one. . One is the compactiﬁcation on a torus T 7 . this implies that the physical value of Newton’s constant. . We denote the compactiﬁed coordinate by x10 which now parameterizes a circle of length L13 . . The ﬁelds are thus decomposed in a Fourier series as periodic functions in x10 on the interval 0 ≤ x10 ≤ L. It is a key element in ensuring that solutions of the lowerdimensional theory remain solutions of the original higher-dimensional one. does not necessarily coincide with the parameter κ2 in the Lagrangian but it also D depends on the precise value adopted for the (ﬂat) metric in the ground state of the theory. mainly in the context of the reduction of D = 11 supergravity to D = 10 dimensions.

ν label the 10-dimensional coordinates and the factor multiplying φ is for convenience later on. we have x10 → x10 − ξ(x) and xµ → xµ . 50]) and an extensive discussion can be found in [18]. In the previous section we noted the existence of certain scale transformations of the D = 11 ﬁelds. Because these originate from supermultiplets that are massless in higher dimensions. is a standard feature of Kaluza-Klein theories [48]. In the compactiﬁed situation we can also involve the compactiﬁcation length into the dimensional scaling. (3. leading to Vµ (x) → Vµ (x) + ∂µ ξ(x) . Following the discussion in [49] we distinguish between symmetries that have a direct explanation in terms of the higher dimensional symmetries.6. Here the x10 -independent component of Vµ acts as a gauge ﬁeld associated with reparametrizations of the circle coordinate x10 with an arbitrary function ξ(x) of the 10 remaining spacetime coordinates xµ . Another point of interest concerns the nature of the massive supermultiplets. e.30) where the indices µ. . [9. which did not leave the theory invariant but could be used to adjust the coupling constant κ10 . which deals with isometries in N = 2 supersymmetric Maxwell-Einstein theories in D = 5. The emergence of new internal symmetries in theories that originate from a higher-dimensional setting. Our derivation here was alluded to in [49].B. Implications of these BPS supermultiplets will be discussed in more detail in Section 3. The 11-dimensional metric can be decomposed according to ds2 = gµν dxµ dxν + e4φ/3 (dx10 + Vµ dxµ )(dx10 + Vν dxν ) . they are 1/2-BPS multiplets which are shortened by the presence of central charges corresponding to the momenta in the compactiﬁed dimension. which correspond to the nontrivial Fourier modes in x10 . Let us start with the symmetries associated with the metric tensor. The massless modes correspond to the x10 -independent parts of the 10-dimensional metric gµν . the vector ﬁeld Vµ and the scalar φ. The integration over x11 introduces an overall factor L in the action (we do not incorporate any L-dependent normalizations in the 14 There are various discussions of this symmetry in the literature. and symmetries whose origin is obscure from a higher-dimensional viewpoint. Speciﬁcally. couple to this gauge ﬁeld with a charge that is a multiple of eKK = 2π · L (3. Its existence in 10-dimensional supergravity was noted long ago (see.31) The massive modes.g. (3.32) Another symmetry of the lower-dimensional theory is more subtle to identify14 . 4 and 3 dimensions. de Wit: Supergravity 39 massless states.

34) L → e−9α L .40 Unity from Duality: Gravity.37) The presence of the above scale symmetry is conﬁrmed by the resulting 10dimensional Lagrangian for the massless (i. 15 Note that these transformations apply uniformly to all Fourier modes. the scale transformations (3. Therefore. (3. This issue will be relevant in Section 3. (3. It simply expresses the length of the x10 -periodicity interval. corresponding to a reparametrization of the 11-th coordinate.6. C11µν → e6α C11µν . (3.36) The tensor gauge ﬁeld CMN P decomposes into a 3.38) .26) are thus suitably combined with the diﬀeomorphism (3. Stated diﬀerently. (3. because the Kaluza-Klein charges (3. these combined transformations are given by15 ea → e−α ea .35) so that κ10 remains invariant. Vµ → e−9α Vµ . µ µ φ → φ + 12α . L itself has no intrinsic meaning. x10 → e−9α x10 . as those depend on x10 /L which is insensitive to the scale transformation.33) and is of dimension [mass]8 . In the eﬀective 10-dimensional theory. we may rescale L according to (3.32) depend explicitly on L. For the ﬁelds corresponding to the 11-dimensional metric.e. ω) − 1 e e2φ (∂µ Vν − ∂ν Vµ )2 2 8 κ2 10 1 − 48 e e2φ/3 (Fµνρσ )2 − 3 e e−2φ/3 (Hµνρ )2 4 √ µ −µ 1 + 1152 2 ε 1 10 C11µ1 µ2 Fµ3 µ4 µ5 µ6 Fµ7 µ8 µ9 µ10 . However. which depends on the coordinatization. Its purely bosonic terms read L10 = 1 − 1 e e2φ/3 R(e. Gauge Theory and Strings Fourier sums. which transform according to Cµνρ → e−3α Cµνρ . This does not imply that the Lagrangian remains invariant when retaining the higher Fourier modes. the coupling constant that emerges in the 10-dimensional theory equals 1 L 2 = κ2 . as long as we change L accordingly. so that the 10-dimensional and the 11-dimensional ﬁelds are directly proportional). Consequently we are then dealing with a symmetry of the Lagrangian. because of the invariance under diffeomorphisms.and a 2-rank tensor in 10 dimensions. κ10 11 (3.35) to yield an invariance of the Lagrangian. x10 -independent) modes. In particular.. we can reparameterize x10 by some diﬀeomorphism.

According to the above. The above example exhibits many of the characteristic features of dimensional reduction and of the symmetries that emerge as a result. n). There are thus 1 2 n(n − 1) independent shift transformations. Therefore the manifold is homogeneous (for a discussion of such manifolds. As already explained. and which combine with the previous ones to generate the group SO(n.6). which act transitively on the manifold (i. i. and n2 scalar ﬁelds. The scalar ﬁelds originate from the metric and the antisymmetric gauge ﬁeld with both indices taking values in T n . 2 they leave no point on the manifold invariant). Before continuing our general discussion. see Sect. 2n Abelian vector gauge ﬁelds. let us brieﬂy discuss an example of the latter based on gravity coupled to an antisymmetric tensor gauge ﬁeld in D + n spacetime dimensions. we have now identiﬁed 1 n(3n − 1) isometries. in total. 3. When reducing to lower dimension one can follow the same procedure a number of times. 2 4 (3. The homogeneous space can then be identiﬁed as the coset space SO(n. act on gij and Bij according to g → OT g O and B → OT B O.B. de Wit: Supergravity 41 where Hµνρ = 6 ∂[µ Cνρ]11 is the ﬁeld strength tensor belonging to the 2-rank tensor gauge ﬁeld. which can be regarded as a generalization of the scale transformations (3.e. 4.36) and (3.39) After compactiﬁcation on a torus T n . the ﬁelds that are independent of the torus coordinates remain massless ﬁelds in D dimensions: the graviton. which exhibits a number of invariances that ﬁnd their origin in the diﬀeomorphisms and gauge transformations related to the torus coordinates.e. all the scalars that emerge from dimensional reduction of gauge ﬁelds are subject to constant shift . L ∝ − 1 E R − 9 E (∂[M BN P ] )2 . n)/(SO(n) × SO(n)). so that.. consecutively reducing the dimension by unit steps. The diﬀeomorphisms acting on the torus coordinates xi which are linear in xi . whose origin is 2 not directly related to the higher-dimensional context. Special tensor gauge transformations with parameters proportional to Λij xj induce a shift of the massless scalars Bij proportional to the constants Λ[ij] . or one can reduce at once to lower dimensions. it turns out that there exist 1 n(n − 1) additional isometries. Furthermore. one must properly account for the periodicity intervals of the torus coordinates xi . so that they are parametrized by a symmetric tensor gij and an antisymmetric tensor Bij . The group GL(n) contains the rotation group SO(n). 11-dimensional supergravity reduced on a hypertorus thus leads to a Lagrangian for the massless sector in lower dimensions (the massive sector is discussed in Sect. its remaining part depends on 1 n(n + 1) parameters.37). xi → Oi j xj . The matrices O generate the group GL(n). However. one tensor gauge ﬁeld. but the action for the massless ﬁelds remains invariant under continuous GL(n) transformations.1). 2 exactly equal to the number of independent ﬁelds gij .

The reason for this is that the isometries may act by means of duality transformations on ﬁeld strengths associated with antisymmetric tensor gauge ﬁelds of rank 1 D − 1 which cannot be implemented on 2 the gauge ﬁelds themselves. Of course. i. for D = 6 we refer to [52]. A more recent discussion of these isometry groups from the perspectives of string theory and M-theory can be found in. because the maximal supergravity theory that one obtains from compactiﬁcation on a hypertorus has no additional coupling constants (beyond Newton’s constant) which could induce R-symmetry breaking. Gauge Theory and Strings transformations. another (noncompact) subgroup is the group GL(11 − D). n) is the T -duality group. for instance. 50]. which emerges for D < 10 can be understood within the string perspective. associated with the reduction on an (11 − D)-dimensional torus. and thus to the number of compactiﬁed dimensions. Usually a counting argument (of the type ﬁrst used in [53]) then readily indicates what the structure is of the corresponding homogeneous space that is parametrized by the scalar ﬁelds. It also follows from the toroidal compactiﬁcations of IIB supergravity. For supergravity. The group SO(n. these subalgebras will partly overlap. One of them is that HR is always the maximal compact subgroup of G. Earlier versions of such tables can. see [51]). In general these extra symmetries are not necessarily symmetries of the full action. Here we merely stress a number of characteristic features of the group G.e. Since 11-dimensional supergravity has itself no scalar ﬁelds.. which has a manifest SL(2) in D = 10 dimensions. This group. SL(2) is the S-duality group and SO(n. 54]. the rank of the resulting symmetry group in lower dimensions is equal to the rank of GL(n). These scalars and the scalars originating from the metric transform transitively under the isometry group. the symmetries may not leave the Lagrangian. We expect that HR is also realized as a symmetry. n) is associated with . In 4 dimensions this phenomenon is known as electric-magnetic duality (for a recent review. with an isometry group whose generators belong to a solvable subalgebra associated with the shift transformations. Therefore we expect that the target space for the scalar ﬁelds is an homogeneous space. where D is the spacetime dimension to which we reduce. for example [24. because the latter do not yet incorporate the full R-symmetry group of the underlying supermultiplet. Yet another subgroup is SL(2) × SO(n. be found in [9. In Table 12 we list the isometry group G and the isotropy group HR of these scalar manifolds for maximal supergravity in dimensions 3 ≤ D ≤ 11. r = 11 − D. but only the ﬁeld equations. In even dimensions. invariant. As we mentioned already. where n = 10 − D.42 Unity from Duality: Gravity. n). it is easy to see that the scalar manifold (as well as the rest of the theory) possesses additional symmetries beyond the ones that follow from higher dimensions. to the subalgebra of GL(11 − D) scale transformations and/or to the subalgebra associated with HR .

Here we should add that it is generally possible to realize the group HR as a local symmetry of the Lagrangian. but only under the local HR group. The ﬁelds which initially transform only under the local HR group. which transform under the group HR as we have shown in Table 8. will now transform under the duality group G through ﬁeld-dependent HR transformations. as this would be in conﬂict with their own gauge invariance. 3. The corresponding connections are then composite connections.39)). In such a formulation most ﬁelds (in particular. de Wit: Supergravity 43 Table 12. 1)/Z2 SL(2) GL(2) E3(+3) ∼ SL(3)×SL(2) E4(+4) ∼ SL(5) E5(+5) ∼ SO(5. We discuss some of the consequences for the central charges and the BPS states in Section 3. This does not pose a serious problem. the G-transformations become realized nonlinearly (we discuss such nonlinear realizations in detail in Sects. the gauge ﬁelds cannot transform under the local group HR . The type-IIB theory cannot be obtained from reduction of 11-dimensional supergravity and is included for completeness. governed by the Cartan-Maurer equations.6. Homogeneous scalar manifolds G/H for maximal supergravities in various dimensions. (3. After ﬁxing a gauge. the fermions) do not transform under the group G.38) does not contain the standard Einstein-Hilbert term for gravity. 5) E6(+6) E7(+7) E8(+8) H 1 1 SO(2) SO(2) U (2) U Sp(4) U Sp(4)×U Sp(4) U Sp(8) SU (8) SO(16) dim [G] − dim [H] 0−0=0 1−0=1 3−1=2 4−1=3 11 − 4 = 7 24 − 10 = 14 45 − 20 = 25 78 − 36 = 42 133 − 63 = 70 248 − 120 = 128 the invariance of toroidal compactiﬁcations that involve the metric and an antisymmetric tensor ﬁeld (cf. while a standard kinetic term for the scalar ﬁeld φ is lacking. This phenomenon is also realized for the central charges. listed in Table 10. D 11 10A 10B 9 8 7 6 5 4 3 G 1 SO(1. The scalars transform linearly under both the rigid duality group as well as under the local HR group. The diﬀerence of the dimensions of G and H equals the number of scalar ﬁelds.B. In this form the gravitational ﬁeld and the scalar ﬁeld are entangled and one has to deal with the scalar-graviton .5 Frames and ﬁeld redeﬁnitions The Lagrangian (3. 4 and 5).

In the case that we include the massive modes. (3. . 16 Note that under a local scale transformation ea → eΛ ea . The corresponding Lagrangian reads16 LEinstein = 10 1 e − 1 R(e. Observe that gauge ﬁelds cannot be redeﬁned by these local scale transformations because this would interfere with their own gauge invariance. which in the metric speciﬁed by (3. This frame is characterized by a standard Einstein-Hilbert term and by a graviton ﬁeld that is invariant under the scale transformations (3.30) is equal to L exp[2 φ /3]. In the context of Kaluza-Klein theory this factor is known as the “warp factor”. After applying the ﬁrst rescaling (3. 3. ω) − 1 (∂µ φ)2 − 1 e e3φ/2 (∂µ Vν − ∂ν Vµ )2 2 4 8 κ2 10 3 1 − 4 e e−φ (Hµνρ )2 − 48 e eφ/2 (Fµνρσ )2 √ 1 + 1152 2 εµ1 −µ10 C11µ1 µ2 Fµ3 µ4 µ5 µ6 Fµ7 µ8 µ9 µ10 . µ µ ea = e−φ/3 [ea ]string . which lead to the so-called Einstein and to the string frame. To separate the scalar and gravitational degrees of freedom.44 Unity from Duality: Gravity.40) We already stressed that the the compactiﬁcation length L is just a parameter length with no intrinsic meaning as a result of the fact that one can always apply general coordinate transformations which involve x10 .36. this rescaling may depend on the extra coordinate x10 .41) Supergravity theories are usually formulated in this frame. respectively. one applies a so-called Weyl rescaling of the metric gµν by an appropriate function of φ. one may also consider the geodesic length. Of course. In the Einstein frame.40) to the Lagrangian (3. the geodesic length of the 11-th dimension is invariant under the SO(1.38) one obtains the Lagrangian in the Einstein frame. Gauge Theory and Strings system as a whole. µ µ (3. where the isometries of the scalar ﬁelds do not act on the graviton.37). They are deﬁned by ea = e−φ/12 [ea ]Einstein . the Ricci scalar in D µ µ dimensions changes according to R → e−2Λ R + 2(D − 1)D µ ∂µ Λ + (D − 1)(D − 2)g µν ∂µ Λ ∂ν Λ . 1) transformations. For these lectures two diﬀerent Weyl rescalings are particularly relevant.

equivalently.55].32). as in (3.43) Let us now return to 11-dimensional supergravity with the 11-th coordinate compactiﬁed to a circle so that 0 ≤ x10 ≤ L. For later purposes let us note that the above discussion can be generalized to arbitrary spacetime dimensions. The metric gµν . ω) + 2(∂µ φ)2 − 3 (Hµνρ )2 2 4 κ2 10 1 − 1 e (∂µ Vν − ∂ν Vµ )2 − 48 e (Fµνρσ )2 8 √ 1 + 1152 2 εµ1 −µ10 C11µ1 µ2 Fµ3 µ4 µ5 µ6 Fµ7 µ8 µ9 µ10 .42). The Einstein frame in any dimension is deﬁned by a gravitational action that is just the Einstein-Hilbert action. (3.B. whereas in the string frame the Ricci scalar is multiplied by a dilaton term exp(−2φD ).41) and (3. L itself has no intrinsic meaning and it is better to consider the geodesic radius of the 11-th dimension.40) leads to the Lagrangian in the string frame.30).44) This result applies to the frame speciﬁed by the 11-dimensional theory17 . In particular the Kaluza-Klein gauge ﬁeld Vµ corresponds in the string context to the R-R gauge ﬁeld of type-II string theory. states with R-R charge are solitonic. 17 This is the frame speciﬁed by the metric given in (3. respectively. which leads to the Lagrangian (3. R-R charges are carried by the D-brane states. which reads R10 = L 2 e 2π φ /3 . deﬁned in (3.38). In the context of 10-dimensional supergravity. In string theory. de Wit: Supergravity 45 The second rescaling (3. The Weyl rescaling which connects the two frames is given by. The signiﬁcance of the dilaton factors in the Lagrangian above is well known. On the other hand the vector Vµ and the 3-form Cµνρ describe Ramond-Ramond (R-R) states and the speciﬁc form of their vertex operators forbids any tree-level coupling to the dilaton [18. the antisymmetric tensor Cµν11 and the dilaton φ always arise in the Neveu-Schwarz sector and couple universally to e−2φ . As we stressed already. (3.37). µ µ (3. or.42) This frame is characterized by the fact that R and (Hµνρ )2 have the same coupling to the scalar φ. In string theory φ coincides with the dilaton ﬁeld that couples to the topology of the worldsheet and whose vacuum-expectation value deﬁnes the string coupling constant according to gs = exp( φ ). that gµν and C11µν transform with equal weights under the scale transformations (3. [ea ]string = e2φD /(D−2) [ea ]Einstein . Lstring = 10 1 e e−2φ − 1 R(e.36. 3. . The inﬁnite tower of massive KaluzaKlein states carry a charge quantized in units of eKK .

M E10 ∂M = e−2φ/3 ∂10 . However. this result becomes (M KK )string = |eKK | e− φ · (3. (3. the KaluzaKlein masses are multiples of M KK = 1 · R10 (3. which is proportional to the Kaluza-Klein charge. meaning that they are contained in supermultiplets that are “shorter” than the generic massive supermultiplets because of nontrivial central charges. 3.46 Unity from Duality: Gravity. From the 11-dimensional expressions.48) M KK = |eKK | e−2 φ /3 . There one has nonperturbative (in the string coupling constant) states which carry R-R charges. In the string frame. Observe that L is ﬁxed in terms of κ10 and κ11 (cf. The most conspicuous change is that the continuous nonlinearly realized symmetry group G .47) Hence Kaluza-Klein states have a mass and a Kaluza-Klein charge (cf. the above result reads (R10 )string = L e 2π φ . (3.32)) related by (3. Gauge Theory and Strings In the string frame.33)).45) It shows that a small 11-th dimension corresponds to small values of exp φ which in turn corresponds to a weakly coupled string theory. (3. The presence of these BPS states introduces a number of qualitative changes to the theory which we discuss in this section. when compactifying dimensions one also encounters massive Kaluza-Klein states. in the frame speciﬁed by the 11-dimensional theory.6 Kaluza–Klein states and BPS-extended supergravity In most of this section we restrict ourselves to pure supergravity. (3. M µ Ea ∂M = ea (∂µ − Vµ ∂10 ) .46) where a and µ refer to the 10-dimensional Lorentz and world indices. The surprising insight that emerged. On the supergravity side these states often appear as solitons. we infer that. is that the Kaluza-Klein features of 11-dimensional supergravity have a precise counterpart in string theory [54–56]. which couple to the supergravity theory as massive matter supermultiplets.49) Massive Kaluza-Klein states are always BPS states. The central charge here is just the 10-th component of the momentum.

e. Similar observations exist for stringlike and membranelike central charges except that in these cases the dimension must be restricted even further [41]. This is the reason why retaining the KaluzaKlein states in the dimensional compactiﬁcation will lead to a breaking of the U -duality group. via ﬁeld-dependent HR transformations). Therefore nonzero central charges must couple to appropriate gauge ﬁelds in the supergravity theory. transform nonlinearly under G. This U -duality group has been conjectured to be the exact symmetry group of (toroidally compactiﬁed) M-theory [54]. in view of the fact that they appear in the anticommutator of two supercharges and supersymmetry is realized locally. upon gauge ﬁxing. one may thus envisage a (possibly local ﬁeld) theory of BPS states coupled to supergravity that is U -duality invariant. Here one naively assumes that the U -duality group acts on the central charges of the BPS states and it is simply deﬁned as the arithmetic subgroup of G that leaves the central-charge lattice invariant. these HR representations can be elevated to representations of G. The BPS states (which are contained in M-theory) should therefore be assigned to representations of the U -duality group.. In most cases. 4 and 5). Another aspect of the coupling of the BPS states to supergravity is that the central charges should be related to local symmetries. This theory would exhibit many of the features of M-theory and describe many of the relevant degrees of freedom. Inspection of the tables that we have presented earlier. From the eleven-dimensional perspective it is easy to see why the central charges associated with the Kaluza-Klein states are too restricted. shows that the gauge ﬁelds usually appear in the G-representation required for gauging the corresponding central charge. However. because the central charges that they carry are too restricted. These gauge ﬁelds transform (with minor exceptions) linearly with constant matrices under the group G. they do not constitute representations of the U -duality group. known as the U -duality group. The Kaluza-Klein states that we encounter in toroidal compactiﬁcations of supergravity are a subset of the 1/2-BPS states in M-theory. i. They carry pointlike central charges and they couple to the Kaluza-Klein photon ﬁelds. because under U -duality the central charges related to the . Provided that the central charges can be assigned to the appropriate representations of the U -duality group and that the appropriate gauge ﬁelds are available. by multiplying with the representatives of the coset space G/HR (representatives of coset spaces will be discussed in Sects. central charges are in principle assigned to representations of the group HR and not of the group G (although the central charges will eventually. the pointlike (ﬁeld-dependent) central charges can be assigned to representations of G for spacetime dimensions D ≥ 4.B. In this way. the vector gauge ﬁelds that emerge from the higher-dimensional metric upon the toroidal compactiﬁcation. However. de Wit: Supergravity 47 is broken to an arithmetic subgroup.

Such extended theories are called BPS-extended supergravity [40.48 Unity from Duality: Gravity. (3.3 are the real symmetric Pauli matrices. Nevertheless. We note that the central charge associated with the parameter a transforms as a doublet under the . In 9 spacetime dimensions with N = 2 supersymmetry the Lorentzinvariant central charges are encoded in a two-by-two real symmetric matrix Z ij . In 9 dimensions the R-symmetry group and the duality group are equal to HR = SO(2) and G = SO(1. The corresponding interacting ﬁeld theory is the unique N = 2 supergravity theory in 9 spacetime dimensions. R). although the theory would presumably not be able to decompactify uniformly to a ﬂat spacetime of more than eleven dimensions. It is well known that the massive supermultiplets of IIA and IIB string theory coincide. 41].12)) in order to deﬁne representations of the U -duality group. the BPS supermultiplets which carry momentum along the circle. has implications for the winding states in order that T -duality remains valid [19]. so that a U -duality invariant theory is obtained. conventional dimensional compactiﬁcation does not involve any brane charges. Henceforth the momentum states of the IIA and the IIB theories will be denoted as KKA and KKB states. string theory and (super)membranes. respectively. IIA and IIB states that are massless in 9 spacetime dimensions.50) Here σ1. However. The fact that they constitute inequivalent supermultiplets. remain inequivalent as they remain assigned to the inequivalent representations of the group SO(8) which is now associated with the restframe (spin) rotations of the massive states. respectively. 1) × SL(2. whereas the massless states comprise inequivalent supermultiplets for the simple reason that they transform according to different representations of the SO(8) helicity group. the number of spacetime dimensions could exceed eleven. We start by considering the BPS multiplets that are relevant in 9 spacetime dimensions from the perspective of supergravity. However. in certain cases one may still be able to extend the Kaluza-Klein states with other BPS states. Gauge Theory and Strings momentum operator in the compactiﬁed dimensions combine with the twoand ﬁve-brane charges (cf. the charges carried by the Kaluza-Klein states) implies that the newly introduced states (associated with wrapped branes) may also have an interpretation in terms of extra dimensions. transform according to identical representations of the SO(7) helicity group and constitute equivalent supermultiplets.e. which can be decomposed as Z ij = b δ ij + a (cos θ σ3 + sin θ σ1 )ij . The aim of this section is to elucidate some of these ideas in the relatively simple context of N = 2 supergravity in D = 9 spacetime dimensions. The fact that some of the central charges are associated with extra spacetime dimensions (i. (2. In this way. When compactifying the theory on a circle.

according to which there exists a IIA and a IIB perspective. so that either MBPS = |pL | or 2 2 |pR | with pL = pR . de Wit: Supergravity 49 SO(2) R-symmetry group that rotates the two supercharge spinors. where pL = −pR .51) Here one can distinguish three types of BPS supermultiplets. Observe that the 1/4-BPS states will never become massless. 2 2 (3. One type has central charges b = 0 and a = 0. which can be generated by wrapping the membrane over the corresponding T 2 .and right-moving momenta. . so that they don’t play a role in what follows. with decompactiﬁcation radii are that inversely proportional and with an interchange of winding and momentum states.B. The 1/4-BPS multiplets arise for string states that have either right. while IIB winding states constitute KKA supermultiplets. The KKA supermultiplets that comprise Kaluza-Klein states of IIA supergravity are of this type. pR . These are 1/2-BPS muliplets. The KKB supermultiplets that comprise the Kaluza-Klein states of IIB supergravity are of this type.53) For pL = pR we conﬁrm the original identiﬁcation of the momentum states. 17]. Subsequently one shows that BPS states that carry these charges must satisfy the mass formula. For type-II string theory one obtains these central charges in terms of the left. ij 1 (pL + pR )δ ij + 1 (pL − pR )σ3 . (for IIA) L R L R 3 2 2 The corresponding BPS mass formula is thus equal to MBPS = 1 |pL + pR | + 1 |pL − pR | . while IIB momentum states constitute KKB supermultiplets. (3.52) Z = 1 (p − p )δ ij + 1 (p + p )σ ij . For the winding states. Another type of 1/2-BPS mulitplets has central charges a = 0 and b = 0. namely that IIA momentum states constitute KKA supermultiplets. one obtains the opposite result: IIA winding states constitute KKB supermultiplets. because they are annihilated by half of the supercharges. while the central charge proportional to the parameter b is SO(2) invariant. the result takes a diﬀerent form for the IIA and the IIB theory as the following formula shows. (for IIB) 2 2 ij (3. that characterize winding and momentum along S 1 . Finally there are 1/4-BPS multiplets (annihilated by one fourth of the supercharges) characterized by the fact that neither a nor b vanishes. Assuming that the two-brane charge takes values in the compact coordinates labelled by 9 and 10. pL . MBPS = |a| + |b| . All of this is entirely consistent with T -duality [16. It is also possible to view the central charges from the perspective of the 11-dimensional (super)membrane [57]. However.or left-moving oscillator states.

This suggests to consider N = 2 supergravity in 9 spacetime dimensions and couple it to the simplest BPS supermultiplets corresponding to KKA and KKB states. the Kaluza-Klein states transform as KKA multiplets. Therefore there is a rather natural way to describe the IIA and IIB momentum and winding states starting from a (super)membrane in eleven spacetime dimensions. as well as eleven-dimensional supergravity.2 denote the momentum numbers on the torus and p is the number of times the membrane is wrapped over the torus. In the case at hand this new theory describes the ten-dimensional IIA and IIB theories in certain decompactiﬁcation limits. Here we have a “single” tower of KKB supermultiplets. Their charges transform obviously with respect to an SO(2) associated with rotations of the coordinates labelled by 9 and 10.54) When compactifying on a torus with modular parameter τ and area A. ij ij Z ij = Z9 10 δ ij − (P9 σ3 − P10 σ1 ) . Hence we have a “double” tower of these charges with corresponding KKA supermultiplets. from the perspective of IIB compactiﬁed on S 1 . As shown in Tables 8 and 10 there are three central charges and 9-dimensional supergravity possesses precisely three gauge ﬁelds that couple to these charges. from the perspective of 9-dimensional supergravity one is led to couple both towers of KKA and KKB supermultiplets simultaneously. Gauge Theory and Strings one readily ﬁnds the expression. On the other hand. This point was ﬁrst emphasized in [58]. not even in a uniform decompactiﬁcation limit. Tm denotes the supermembrane tension. In the case at hand we know a lot about these couplings from our knowledge of the T 2 compactiﬁcation of D = 11 supergravity and the S 1 compactiﬁcation of IIB supergravity. is not known. as the ﬁelds never depend on all the 12 coordinates! Whether this kind of BPS-extended supergravity oﬀers a viable scheme in a more general context than the one we discuss here. the BPS mass formula takes the form MBPS = = 2 P92 + P10 + |Z9 10 | 1 √ |q1 + τ q2 | + Tm A |p| . although there is no 12-dimensional Lorentz invariance. Clearly the KKA states correspond to the momentum modes on T 2 while the KKB states are associated with the wrapped membranes on the torus. the Kaluza-Klein states constitute KKB multiplets and their charge is SO(2) invariant. From the perspective of 11-dimensional supergravity compactiﬁed on T 2 . In that case one obtains some dichotomic theory [19]. But the theory is in some sense truly 12-dimensional with three compact coordinates. However. which we refer to as BPS-extended supergravity. . A τ2 (3.55) Here q1.50 Unity from Duality: Gravity. (3.

we stress that. 2). as we stressed above. which are described by a nonlinear sigma model based on SL(2. which satisfy a constraint φα φα = 1 and are subject to a local SO(2) invariance. which is equal to exp(− 4 σ). The scalar ﬁeld σ is related to G99 . From the perspective of the supermembrane. The resulting BPS-extended theory incorporates 11-dimensional supergravity and the two type-II supergravities in special decompactiﬁcation limits. The precise relationship follows from comparing the SO(1. In 9 dimensions supergravity has two more scalars. denoted by Bµ . The coset is described by the complex doublet of ﬁelds φα . There are two vector ﬁelds Aα . de Wit: Supergravity 51 The ﬁelds of 9-dimensional N = 2 supergravity are listed in Table 13. which is consistent with the fact that the IIA winding states constitute KKB supermultiplets. a symmetry of the massless theory that emerges upon dimensional reduction and is associated with scalings of the internal vielbeine. where we also indicate their relation with the ﬁelds of 11-dimensional and 10-dimensional IIA/B supergravity upon dimensional reduction. which are the µ Kaluza-Klein photons from the T 2 reduction of eleven-dimensional supergravity and which couple therefore to the KKA states. the IIB metric component in the compactiﬁed dimension. 1). But. which conﬁrms that they couple to the IIB (elementary and D1) winding states. so that they describe precisely two scalar degrees of freedom (α = 1. These two ﬁelds transform under SL(2). It is not necessary to work out all the nonlinear ﬁeld redeﬁnitions here. while the KKB states correspond to the membranes wrapped around the torus.B. We already mentioned the three Abelian vector gauge ﬁelds which couple to the central charges. the KKA states are the momentum states on T 2 . We expect that the local SO(2) invariance can be incorporated in the full BPS-extended supergravity theory and can be exploited in the construction of the couplings of the various BPS supermultiplets to supergravity. by G99 = exp(σ). so that it couples to the KKB states. While it is gratifying to see how all these correspondences work out. from the perspective of 9-dimensional N = 2 supergravity. likewise it is related to the determinant of the 11-dimensional metric in the compactiﬁed dimensions. the results follow entirely from supersymmetry. The third gauge ﬁeld. On the IIA side it originates from the IIA tensor ﬁeld. R)/SO(2). From the IIB side they originate from the tensor ﬁelds. we are dealing with a 12-dimensional theory . is a singlet under SL(2) and is the KaluzaKlein photon on the IIB side. which can be understood from the perspective of the modular transformation on T 2 as well as from the S-duality transformations that rotate the elementary strings with the D1 strings. From the IIA perspective these correspond to the Kaluza-Klein states on S 1 and the D0 states. as the corresponding ﬁelds can be uniquely identiﬁed by their scaling weights under SO(1. 1) 3 weights through the dimensional reduction of IIB and eleven-dimensional supergravity.

In this way we are able to deduce the following BPS mass formula. The last column lists the SO(1. respectively. We return to them in a moment.56) where qα and p refer to the integer-valued KKA and KKB charges. (3. Z). .52 Unity from Duality: Gravity. Cµν Cµνρ φ. type-IIA. G10 10 IIA Gµν Cµ 9 Gµ 9 . are split as M = (µ. . The 11-dimensional and ˆ 10-dimensional indices. G9 9 . One simply has the freedom to view the theory from a IIA or a IIB perspective and interpret it accordingly. This formula can be compared to the membrane BPS formula (3. where µ = 0. Aµν 10 ˆ Aµνρ ˆ ˆ ˆ G9 10 . C9 D=9 gµν Bµ α Aµ α Aµν IIB Gµν Gµ 9 α Aµ 9 α Aµν SO(1. we only have a discrete subgroup that leaves the charge lattice invariant. although no ﬁeld can depend nontrivially on all of these coordinates. 1) × SL(2.and T -duality are manifest. D = 11 ˆ Gµν ˆ Aµ 9 10 ˆ ˆ Gµ 9 . G9 9 . 9). 8. Gµ 10 ˆ ˆ Aµν 9 . Both S. Hence. R) of pure supergravity after coupling the theory to the BPS multiplets. 1) 0 −4 3 −1 2 0 7 Aµνρ φ exp(σ) α Aµνρσ φ G9 9 α here. The theory has obviously two mass scales associated with the KKA and KKB states. The bosonic ﬁelds of the eleven dimensional. respectively. 1) scaling weights of the ﬁelds. KKA (3. We should discuss the fate of the group G = SO(1. 1. Gauge Theory and Strings Table 13.57) . The KKA and KKB states and their interactions with the massless theory can be understood from the perspective of compactiﬁed 11-dimensional and IIB supergravity. Cµ Cµν 9 . MBPS (q1 . 10) and M = (µ. and mKKA and mKKB are two independent mass scales. p) = mKKA e3σ/7 |qα φα | + mKKB e−4σ/7 |p| . which is aﬀected by this group. after coupling to the BPS states. The central charges of the BPS states form a discrete lattice. ninedimensional N = 2 and type-IIB supergravity theories. q2 . This is the group SL(2. although the latter has become trivial as the theory is not based on a speciﬁc IIA or IIB perspective.55) in the 11-dimensional frame. One then ﬁnds that m2 mKKB ∝ Tm . . 9.

there exists no theory with Q = 32 supercharges in ﬂat Minkowski spacetime of dimensions D > 11. de Wit: Supergravity 53 Table 14. This multiplet comprises 8 + 8 physical degrees of freedom. D 10 9 8 7 6 5 4 3 HR 1 1 U (1) U Sp(2) U Sp(2)×U Sp(2) U Sp(4) U (4) SO(8) Aµ 1 1 1 1 1 1 1 φ 0 1 1+¯ 1 3 (2. unlike in the case of maximal supergravity. All supermultiplets contain a gauge ﬁeld Aµ . there is no uniform decompactiﬁcation limit. possibly associated with some non-Abelian gauge group. 3. We also consider the Q = 16 supergravity theories. Field content for maximal super-Maxwell theories in various dimensions. Therefore. 1) + (1. Table 14 shows the ﬁeld representations for the vector multiplet in dimension 3 ≤ D ≤ 10.B. For Q = 16 the matter will be in the form of vector supermultiplets. Table 15 summarizes Q = 16 supergravity for dimensions 3 ≤ D ≤ 10. the most important conclusion to draw from (3.58) . 2) 4 4+¯ 4 8 with a numerical proportionality constant.38). In D = 3 dimensions the vector ﬁeld is dual to a scalar. The Lagrangian can be obtained by truncation of (3. scalars φ and spinors χ and comprises 8 + 8 degrees of freedom. L10 = 1 − 1 e e2φ/3 R(e. However. we now have the option of introducing additional matter ﬁelds. The 6∗ representation of SU (4) is a selfdual rank-2 tensor. However. which are now restricted to dimensions D ≤ 10.56) is that there is no limit in which the masses of both KKA and KKB states will tend to zero. in spite of the fact that we have more than 11 dimensions. 4 (3. In other words. In D = 10 dimensions the bosonic terms of the supergravity Lagrangian take the form [59]. 2) 5 6∗ 8 χ 1 1 1+¯ 1 2 (2.7 Nonmaximal supersymmetry: Q = 16 For completeness we also summarize a number of results on nonmaximal supersymmetric theories with Q = 16 supercharges. ω) 2 κ2 10 3 − 4 e e−2φ/3 (Hµνρ )2 − 1 e(∂µ Aν − ∂ν Aµ )2 .

Aµ → e3α Aµ (3.36. A feature that deserves to be mentioned. 1) 5∗ where. the SU (4) transformations cannot be implemented on the vector potentials. In D = 4 dimensions.37). 1) and (2. . representing an Abelian vector supermultiplet. Note also that the kinetic term for the Kaluza-Klein vector ﬁeld in (3.38). but act on the (Abelian) ﬁeld strengths by duality transformations. The scalars parametrize the coset space SO(8. where Aµ transforms diﬀerently from the Kaluza-Klein vector ﬁeld Vµ . Chern-Simons terms play an important role in the anomaly cancellations of this theory. µ µ 6α φ → φ + 12α . ea → e−α ea . D 10 9 8 7 6A 6B 5 4 3 HR 1 1 U (1) U Sp(2) U Sp(2)×U Sp(2) U Sp(4) U Sp(4) U (4) SO(8) # 64 56 48 40 32 24 24 16 8k graviton 1 1 1 1 1 1 1 1 1 p = −1 1 1 1 1 1 1 1+¯ 1 8k p=0 1 1+¯ 1 3 (2.59) C11µν → e C11µν . 0) supergravity. where k is an arbitrary integer. Gauge Theory and Strings Table 15. with the exception of the 6B and the 4-dimensional theory. Note that. 3. This reﬂects itself in the extension of the symmetry transformations noted in (3. 2) 5+1 [6] p=1 1 1 1 1 (1. depends on φ. In 6 dimensions type-A and type-B correspond to (1. k)/(SO(8)×SO(k)). The tensor ﬁeld in the 6B theory is selfdual. we have included a single vector gauge ﬁeld Aµ . In D = 3 dimensions supergravity is a topological theory and can be coupled to scalars and spinors. is that the ﬁeld strength Hµνρ associated with the 2-rank gauge ﬁeld contains a Chern-Simons term A[µ ∂ν Aρ] .54 Unity from Duality: Gravity. # denotes the number of bosonic degrees of freedom. In the third column. for convenience. unlike the kinetic term for the matter vector ﬁeld in the Lagrangian above. Bosonic ﬁelds of nonmaximal supergravity with Q = 16. all these theories contain precisely one scalar ﬁeld.

namely ea µ ea µ ea µ = e−φ/12 [ea ]Einstein . ω) + 2(∂µ φ)2 2 κ2 10 − 3 e (Hµνρ )2 − 1 e eφ (∂µ Aν − ∂ν Aµ )2 . Lstring = 10 1 −2φ e − 1 e R(e. 4 Homogeneous spaces and nonlinear sigma models This chapter oﬀers an introduction to coset spaces and nonlinear sigma models based on such target spaces with their possible gaugings. The aim . ω) + 2e(∂µ φ)2 2 κ2 10 − 3 e (Hµνρ )2 − 1 e(∂µ Aν − ∂ν Aµ )2 . This is the low-energy eﬀective Lagrangian relevant for the heterotic string.61) 4 The second Weyl rescaling leads to the following Lagrangian. This is related to the fact that the SO(32) heterotic string theory is S-dual to type-I string theory [60]. As it turns out. (3. where the type-I dilaton must be associated with −φ.60) It is straightforward to obtain the corresponding Lagrangians. Finally. 4 4 (3. µ = e−φ/3 [ea ]string .62) which shows a uniform coupling to the dilaton. ω) − 1 e(∂µ φ)2 2 4 κ2 10 3 − 4 e e−φ (Hµνρ )2 − 1 e e−φ/2 (∂µ Aν − ∂ν Aµ )2 . 4 4 (3. the third Weyl rescaling yields Lstring 10 = 1 e e2φ − 1 R(e. µ (3. the graviton is again invariant under the isometries of the scalar ﬁeld.63) Now the dilaton seems to appear with the wrong sign. this is the low-energy eﬀective action of the type-I string. µ = eφ/6 [ea ]string . In the Einstein frame. de Wit: Supergravity 55 In this case there are three diﬀerent Weyl rescalings that are relevant. The bosonic terms read LEinstein = 10 1 − 1 e R(e.B. Eventually the matter gauge ﬁeld has to be part of an non-Abelian gauge theory based on the groups SO(32) or E8 × E8 in order to be anomaly-free.

We start by introducing the concept of a coset space G/H. . 0. concentrating on the maximal supergravities in D = 4. Gauge Theory and Strings of this introduction is to facilitate the discussion in the next chapter.56 Unity from Duality: Gravity. The sphere is obviously invariant under SO(n + 1). these theories have a nonlinearly realized symmetry group. such as [61. equal to E7(7) and E6(6) . 62]. Finally we introduce the so-called gauging of this class of nonlinear sigma models. respectively. A homogeneous space is a space where every two points can be connected by an isometry transformation. . Such invariances are called isometries and G = SO(n + 1) is therefore known as the isometry group.1 Nonlinearly realized symmetries As an example consider the n-dimensional sphere S n of unit radius which we may embed in an (n+1)-dimensional real vector space Rn+1 . Then we discuss the corresponding nonlinear sigma models. which we make use of in later chapters. where H is a subgroup of a group G. denoted by H. The latter enable us to include some material on de Sitter and anti-de Sitter spacetimes. the rotation connecting these two points is not unique as every point on S n is invariant under an SO(n) subgroup. From the north pole. . for a homogeneous manifold. It is convenient to choose a certain point on the sphere (let us call it the north pole) with coordinates in Rn+1 given by (0. based on homogeneous target spaces. where we explain the gauging of maximal supergravity. we can reach each point by a suitable rotation. 4. However. Obviously. In this introduction we try to be as general as possible but in the examples we restrict ourselves to pseudo-orthogonal groups: SO(n) or the noncompact versions SO(p. However. which is a subgroup of these exceptional groups. As discussed earlier. . the isotropy groups for two arbitrary points are isomorphic (but not identical as one has to rotate between these points). Clearly the sphere is such a homogeneous space as every two points on S n can be related by an SO(n + 1) isometry. and present their description in a form that emphasizes a local gauge invariance associated with the group H. the north pole itself is invariant under the SO(n) isotropy group. consisting of the following orthogonal matrices embedded . This group is called the isotropy group (or stability subgroup). 1). q). the group of (n + 1)-dimensional rotations. Most of this material is standard and can be found in textbooks. 5 spacetime dimensions. The construction of these gaugings makes an essential use of the concepts and techniques discussed here. It is possible to elevate the Abelian gauge group associated with the vector gauge ﬁelds to a non-Abelian group.

q) thus satisfy g −1 = η 0 0 1 gT η 0 0 1 . Elements of SO(p. ∗ ··· ∗ 0 ··· 0 0 . . . Every such space can be described in terms of appropriate G/H cosets based on an isometry group G and a isotropy subgroup H. It is not diﬃcult to ﬁnd such a parametrization.2) g(α) = exp −αj 0 with some element of H. 1).B. so that we can also deal with noncompact spaces. For SO(p.1). The noncompact groups leave an indeﬁnite metric invariant. where η is again a diagonal metric with p (or q) eigenvalues equal to −1 and q − 1 (or p − 1) eigenvalues equal to +1.3). and therefore the space S n is a coset space G/H with G = SO(n + 1) and H = SO(n). if a rotation g1 ∈ G. 0 1 57 h∈H. (4. Therefore points on the sphere can be associated with the class of group elements g ∈ G that are equivalent up to multiplication by elements h ∈ H from the right. which assigns a single SO(n + 1) element to every coset.2) satisﬁes the condition (4. . h=. (4. (4. . j = 1. q) we have a diagonal metric with p eigenvalues +1 and q eigenvalues −1 (or vice versa as the overall sign is not relevant). One ﬁrst observes that every element of SO(n + 1) can be decomposed as the product of 0 αi . is thus equivalent to giving a parametrization of the sphere. . A parametrization of the cosets of SO(n + 1)/SO(n). ∗ ··· ∗ . .4) When η equals the unit matrix. we are dealing with a compact space. ∗ . Using the same decomposition as in (4. provided that αi = ηij αj .. we choose this metric of the form diag (η. . n. (4. When η has negative eigenvalues the matrix (4. Here i. Such equivalence classes are called cosets. de Wit: Supergravity into SO(n + 1). The coset representative (4. with G = SO(n + 1) maps the north pole onto a certain point on the sphere. .2) is no longer orthogonal and the . In view of various applications we extend our notation to noncompact versions of the orthogonal group. then the transformation g2 = g1 · h will do the same.1) Therefore. The sphere is obviously just one particular example of a homogeneous space.3) so that the metric η is obviously H-invariant. .

19 We are a little cavalier here with our terminology. 1)/SO(n). However. which ensures that in an inﬁnitesimal neighbourhood of a point invariant under H. we have the anti-de Sitter space SO(n − 1. −. Let us mention some examples. For η = −1 we have the hyperbolic space19 SO(n. In our example. We will not discuss these issues here. This aspect will be important later on when comparing the curvature for these spaces. +) we have the de Sitter space SO(n. 1). We also note that. the latter have been used to generate the coset representative. 1). there are other decompositions than (4. 1)/SO(n − 1. the generators k are deﬁned up to additive terms belonging to elements of the algebra associated with H. but they are pseudo-Riemannian. k] = h. h] = [h.10)). (4.5) The ﬁrst commutation relation states that the h form a subalgebra. Also the de Sitter and anti-de Sitter spaces are hyperbolic. such as in the coset representatives (4. +. +). so that they satisfy (schematically). the homogeneous space is symmetric. Obviously the above relations involve a choice of basis. In this way we can thus treat a variety of spaces at the same time. and for η = (−. which oﬀer distinct advantages. is “mostly plus” for de Sitter. in the noncompact case. k. as we shall see later (cf.2). We reserve the term hyperbolic for the Riemannian hyperbolic space. and “mostly minus” for the hyperbolic and the antide Sitter space. because there are disconnected components. Of course. · · · .2). 18 Noncompact . the coordinates αi rotate under H according to that representation. (4. the coset representative is not unique.58 Unity from Duality: Gravity. The second one implies that the generators k form a representation of H. Gauge Theory and Strings space will be noncompact18 . the latter are associated with the last row and column of (4. Unlike the de Sitter and anti-de Sitter spaces the former spaces are double-sheeded. In the more general case the third commutator may also yield the generators k. which eventually will play the role of the tangent-space metric. note that η. When they do not. In general. k] = [k. have a positive or negative deﬁnite metric and are thus Riemannian spaces. We have decomposed the generators of G into generators h of H and generators k belonging to its complement. for η = (−. · · · .1). h. The elements g(α) ∈ G deﬁne a representative of the G/H coset space. coset spaces where H is the maximal compact subgroup of a noncompact group G. groups are not fully covered by exponentiation. 2)/SO(n − 1. [h. and therefore a parametrization of the corresponding space.

This corresponds to the sphere S n . . because in that case the parameter space can be restricted to 0 ≤ α < π.2). the upper or the lower hemisphere). such as. In all other cases the parameter space is obviously noncompact and the sine and cosine may change to the hyperbolic sine and cosine in those parts of the space where α2 is negative.7) so that the coset representative reads ± 1 − y2 − 1 i i δj + y y j y2 g(y) = −yj (4. (4. Y A = (y i .9) 20 The coordinates (y i . for instance20 .B. i δj + αi αj cos α2− 1 α g(α) = − sin α αj α 59 sin α αi α . For the sphere the range of the coordinates is restricted by y 2 = Σi (y i )2 ≤ 1. Observe that the appearance of η in the above formulae is the result of the fact that the generators k are normalized according to tr(ki kj ) = −2 ηij . We may also chose diﬀerent coordinates. Acting with (4. Note that i the n × n submatrix in (4. the “north pole”) in the (n+1)-dimensional embedding space. Obviously the space is compact when η is positive. 1) yields the following coordinates in the embedding space. Diﬀerent parametrizations are generally related through (coordinate-dependent) H transformations acting from the right. ± 1 − y 2 ) are sometimes called homogeneous coordinates. depending on the sign choice. One may choose a diﬀerent parametrization of the cosets by making a diﬀerent decomposition than in (4. ± 1 − y 2 ) .2). cos α (4.8) equals the square root of the matrix δj − y i yj . 0. Inhomogeneous coordinates are the ratios y i / 1 − y 2 . . de Wit: Supergravity Let us now proceed and evaluate (4. . because the G-transformations act linearly on these coordinates. One may use the coset representative to sweep out the coset space from one point (i.8) ± 1 − y2 where. . we parametrize diﬀerent parts of the space (for the sphere S n .6) where α2 = ηij αi αj . .8) on the point (0. y i = αi sin α .e. α y i (4.

12) We return to this result in the next section. Applying this construction to the case at hand.10) Since the g(y) are contained in G one may examine the eﬀect of G transformations acting on g(y). which changes the coordinates y i by constant rotations. After this multiplication the result is in general no longer compatible with the coset representative g(y). (4. (4. but by applying a suitable y-dependent H transformation. we can again bring g(y) in the desired form. one has g(y) −→ oG g(y) = g(y ) oH (y) . To see this. The inﬁnitesimal transformation y i → y i = ˆ y i + ξ i (y) deﬁnes the so-called Killing vectors ξ i (y). The representation transforms (not necessarily irreducibly) under SO(p. g(α) = exp[ 1 Γi αi ] = cos(α/2) 1 + i 2 sin(α/2) i α Γi . As the reader can easily verify. δy i = i 1 − y2 . q) in the spinor representation. q) generated by the matrices 1 Γij . In other words. both types of transformations take the form of a constant G transformation on the embedding coordinates Y A which leaves the embedding condition (4. q + 2 1). we multiply g(y) by a constant element oG ∈ G from the left. (4. which will induce corresponding transformations in the coset space. Gauge Theory and Strings Using (4. we ﬁnd the relation. α (4.13) where the i are n constant parameters. ˆ g ξ i (y) ∂i g(y) = ˆ g(y) − g(y) h(y) .10) invariant. as one can verify by including extra generators equal to the matrices 1 2 Γi .11) Hence the eﬀect is a change of coordinates y → y in a way that satisﬁes the group multiplication laws. but in fact it transforms also as a spinor under SO(p. One corresponding to the group H. Coset representatives can be deﬁned in diﬀerent representations of the group G. The other corresponds to n coordinate dependent shifts. one ﬁnds that there are two types of isometries. Assume that we have a representation of the Cliﬀord algebra C(p. The most interesting one is the spinor representation.60 Unity from Duality: Gravity. Writing oG ≈ 1 + g ˆ and oH ≈ 1 + h(y). (4. oH (y). q + 1)/SO(p.14) .3) one then shows that the coset space is embedded in n + 1 dimensions according to ηij Y i Y j + (Y n+1 )2 = 1 . from the right. q). Consequently we can deﬁne a representative of SO(p.

Therefore the one-forms can be decomposed into the generators h and k. de Wit: Supergravity 61 √ with α deﬁned as before. speciﬁed at the north pole. The signiﬁcance of this fact will be clear in a sequel. These one-forms e are thus related to the vielbeine of the coset space. Hence e deﬁnes a square matrix.. Γj } = −2ηij 1. Equation (4. so that the one-forms take their value in the Lie algebra associated with G. the one-form g −1 dg takes its values in the Lie algebra associated with G.16) where ω is decomposable into the generators h and e into the generators k. This construction can applied as well to cosets of other (pseudo-)orthogonal groups. g −1 dg = ω + e .15) One can act with this representative on a constant spinor. .17) 21 Observe that in supergravity we have deﬁned the spin connection ﬁeld with opposite sign. In terms of the coordinates y i the representative reads g(y) = 1 2 1+y+ y i Γi √ 1−y 1+ √ · 1+y+ 1−y (4.B. The one-forms g −1 dg are called left-invariant.e. similar results can be obtained in other representations of G. Obviously. As a ﬁrst consequence we note that the spinor ψ(y). where g(y) ∈ G. d + ω + e ψ(y) = 0 . i.but by no means essential. deﬁned with the help of the representative (4. α2 = αi αi . i. Because the g’s themselves are elements of G. The exterior derivative dg(y) describes the change of g induced by an inﬁnitesimal variation of the coset-space coordinates y i . introduced earlier. (4. The one-forms ω deﬁne the spin connection21 . (4. satisﬁes the equation.15) at the end of the previous subsection. with indices i that label the coordinates and indices a that label the generators k.2 Geometrical quantities Geometrical quantities of the homogeneous space are deﬁned from the leftinvariant one-forms g −1 dg. we deﬁne ψ(y) = g −1 (y) ψ(0). because they are invariant under left multiplication of g with constant elements of G.e.16) is of central importance for the geometry of the coset spaces. and {Γi . which deﬁne a tangent frame at each point of the space. We shall exhibit this below. 4. associated with tangent-space rotations that belong to the group H. It is convenient to use the language of diﬀerential forms. The resulting y-dependent spinor ψ(y) is a so-called Killing spinor of the coset space.

ej ] − [ωj . (4.15)). Rij (G/H) = ∂i ej − ∂j ei + [ωi . we note the relations [Di .21) We can also deﬁne Lie-algebra valued curvatures associated with ωi and ei . ωj ] . Dj ] = −Rij (H) .e. Furthermore it follows from (4.22) The values of these curvatures follow from the Cartan-Maurer equations.25) .. (4.19) This equation can again be decomposed (using the ﬁrst two relations (4.16). (4.23) (4.18) This leads to a diﬀerent parametrization of the coset space. (4. yi ωi ei −→ y i + ξ i . which yields ωi ei −→ h−1 ωi h + h−1 ∂i h .e. i. d(ω + e) = −(ω + e) ∧ (ω + e) . (4.62 Unity from Duality: Gravity. Rij (G/H) = Di ej − Dj ei . To derive these equations we take the exterior derivative of the deﬁning relation (4. −→ ωi − ∂i ξ j ωj − ξ j ∂j ωi . i. Let us now proceed and investigate the properties of the one-forms ω and e. ei ] .5) in terms of the generators h and k.20) Obviously. (4. −→ h−1 ei h . It is straightforward to see how ω and e transform under (4. (3. h(y) ∈ H . Rij (H) = ∂i ωj − ∂j ωi + [ωi . or.18) (ω + e) −→ h−1 (ω + e)h + h−1 dh . −→ ei − ∂i ξ j ej − ξ j ∂j ei . Gauge Theory and Strings Upon writing this out in terms of the gamma matrices. one recovers precisely the so-called Killing spinor equation (cf.27) that ωi and ei transform as covariant vectors under coordinate transformations. In general it is not necessary to specify the coset representative. ω acts as a gauge connection for the local H transformations. g(y) −→ g(y) h(y) . Introducing H-covariant derivatives.24) d(g −1 dg) = −(g −1 dg) ∧ (g −1 dg) . as diﬀerent representatives are related by y-dependent H transformation acting from the right on g. (4. in terms of ω and e.

using the same matrix decomposition as before. The quantity ωi thus acts as the connection associated with rotations of the tangent frames. From the . we ﬁnd Rij (H) = −[ei . (4. which follows from the (pseudo)orthogonality of H.5). ab where. In the context of diﬀerential geometry the indices i are called world indices. using the relations (4.29) Furthermore. de Wit: Supergravity 63 Decomposing this equation in terms of the Lie algebra generators. as before. (4. the curvatures introduced before. whereas the indices a. b.28) eia yi y a ± y2 1 − y2 −1 . . Hence. . we ﬁnd explicit expressions for the vielbein and the spin connection.8) one readily obtains ab ωi = = b a y a δi − y b δi a δi + 1∓ 1 1 − y2 y2 . (4.27) g −1 dg = i −ei (y) dy 0 From (4. These aspects are easily recognized in the examples we are discussing. . (4. because the group H was precisely the (pseudo)orthogonal group. The inverse vielbein reads i i e a = δa + ya y i ± y2 1 − y2 − 1 . because they refer to the coordinates of a manifold. indices are raised and lowered with η.5)). ωi (y) dy i ei (y) dy i . the ﬁelds ei can be decomposed into the generators k and thus deﬁne a set of vielbeine eia that specify a tangent frame at each point in the coset space. ej ] . Rij (G/H) = 0 . Because the generators k form a representation of H. and therefore we call it the spin connection. As we already alluded to earlier. this group rotates the tangent frames. b. are readily identiﬁed with the curvature of the spin connection and with the torsion tensor. that label the generators k are called tangent-space indices (or local Lorentz indices in the context of general relativity).26) Note that the vanishing of Rij (G/H) is a consequence of the fact that we assumed that the coset space was symmetric (see the text below (4. Usually the group H can be embedded into SO(n) (or a noncompact version thereof) and leaves some target-space metric invariant (we will see the importance of this fact shortly). Note that ωi is antisymmetric in a.B.

This sign is important when comparing to spheres or de Sitter spaces. there is a more extended class of metrics that one may consider. and we ﬁnd the following result for the Riemann curvature tensor. we thus ﬁnd in components. so that gij = ηab eia ejb . this leads to Γijk = y k gij .34) where gij is the metric tensor deﬁned by (4. (4. a “mostly plus” metric requires to include a minus sign in the deﬁnition (4. (4. .1. (4. the connection coincides with Christoﬀel symbol. Combining (4.31) When there are several H-invariant tensors.31) for the hyperbolic and anti-de Sitter spaces. Because ωi diﬀers in sign as compared to the spin connection used in ab Section 3. (4. explained above.64 Unity from Duality: Gravity. the Riemann tensor is equal to minus the curvature Rij (ω).16).35) ξ i (y) (ωi (y) + ei (y)) = g −1 (y)ˆ g(y) − ˆ . which indicates that we are dealing with a maximally symmetric space.12) with (4. (4.34) is proportional to the metric. which we introduced earlier. ab Rij (ω) = 2 ea eb .31).30) To deﬁne a metric gij one contracts an H-invariant symmetric rank-2 tensor with the vielbeine. as we have discussed above. The Riemann curvature (4.32) Given the fact that we have already made a choice for η previously. c l upon contraction with ηac ek eb . 2 All coset spaces have isometries corresponding to the group G. l l Rijk l = −gki δj + gkj δi . we know that the aﬃne connection is equal to Γijk = eak Di eja . The obvious invariant tensor is ηab . but in the case at hand the metric is unique up to a proportionality factor. from which we can deﬁne the Riemann curvature. The diﬀeomorphisms associated with these isometries are generated by Killing vectors ξ i (y). In the parametrization (4. For the examples at hand. g h(y) (4. From the vielbein postulate. but we stress that this is related to the signature choice that we made for the metric. This means that the maximal number of isometries (equal to 1 n(n + 1)) is realized for this space.33) Because the torsion is zero. gij = ηij + yi yj .28) one obtains for the metric and its inverse. 1 − y2 g ij = η ij − y i y j . Gauge Theory and Strings Cartan-Maurer equations. Thus the curvature is of deﬁnite sign. we obtain. [i j] Di eja − Dj eia = 0 .

36) . we established (cf. β. (4. from which e and ω were constructed.36) yields Rij (G/H) ξ j + [ei . G/H ˆ g −1 (y) g g(y) (4. Hence.11)) that the G-transformation acting on the left can be decomposed into a diﬀeomorphism combined with a coordinate-dependent H-transformation. which transform under H. . .12). Let us label the generators of the group G by indices α. Moreover. the vielbein e and the spin connection ω are invariant under these combined transformations. g ˜ −ξ i (y) ωi (y) + h(y) . ˜(y)]G/H = 0 .39) Combining the ﬁrst equation with the ﬁrst equation (4. Rij (H) ξ j . (4. g g (4. g (4.B.41) Observe that both terms vanish separately for a symmetric space. h(y)] = ˆ ˆ ∂i ξ j ωj + ξ j ∂j ωi + ∂i h(y) + [ωi . . Therefore.42) . we ﬁnd ξ i (y) ei (y) = ˆ h(y) = where ˜ h(y) ˜(y) g = = ˆ g −1 (y) g g(y) H 65 ˜(y) . 0. h(y)] = ˜ In terms of h(y) these results take a more covariant form. it thus follows that the metric is invariant under the diﬀeomorphism associated with G. ˜ ∂i ξ j ej + ξ j Dj ei + [ei . For the vielbein and spin connection. δgij = Di ξj (y) + Dj ξi (y) = 0 . (4. (4. ˆβ ] = fαβγ ˆγ . Since the metric is Hinvariant by construction. . de Wit: Supergravity Decomposing this equation according to the Lie algebra.40) 0. .37) ˆ The contribution h(y) is only relevant for those quantities that live in the tangent space. This follows from (4. h(y)] = ˜ Di h(y) = 0.36). Now we return to the observation that the left-invariant forms. are invariant under the group G. it follows that e and ω are both invariant as well.38) where ξi = gij ξ j and ξ i is the so-called Killing vector deﬁned by (4. (4. The diﬀeomorphisms generated by the Killing vector ﬁelds will give rise to the group G. and introduce structure constants by g [ˆα . Therefore. we ﬁnd ˆ ∂i ξ j ej + ξ j ∂j ei + [ei .

(4. j i j i ξβ ∂j ξα − ξα ∂j ξβ = i fαβγ ξγ . i ˜ −ξα Di + hα .47) The above expressions show that Qµ and Pµ are just the pull backs of the target space connection and vielbein to the spacetime.45) 4. Pµ (φ) = ei (φ) ∂µ φi . one uses the analogue of (4. (4. so that the ﬁelds φi (x) deﬁne a map from the spacetime to the coset space. On such ﬁelds the isometries are generated by the operators.43) One can consider ﬁelds on the coset space. hβ ] = (4. to deﬁne Lie-algebra valued quantities Qµ and Pµ .66 Unity from Duality: Gravity.3 Nonlinear sigma models with homogeneous target space It is now rather straightforward to describe a nonlinear sigma model based on a homogeneous target space by making use of the above framework. which are functions of the coset space coordinates assigned to a representation of the group H. Obviously one has the relations Qµ (φ) = ωi (φ) ∂µ φi . Hence we may follow the same procedure as before and deﬁne a coset representative V(φi (x)) ∈ G. Gauge Theory and Strings The Killing vectors and the corresponding H-transformations then satisfy corresponding group multiplication properties. where Qµ is decomposable into the generators h and Pµ into the generators k. Therefore one may elevate these transformations to transformations that depend arbitrarily on xµ . To show this we note the identity i i i j ˜ ˜ ˜ ˜ (−ξα Di + hα )(−ξβ Di + hβ ) = ξα ξβ (Di Dj − Rij (H)) + hα hβ j i ˜ j ˜ j +[ξα (Di ξβ ) + hα ξβ + hβ ξα ]Dj . The local H transformations depend on the ﬁelds φ(x) and thus indirectly on the spacetime coordinates.44) On the basis of the results above one can show that these operators satisfy the commutation relations of the Lie algebra associated with the isometry group G. (4.16). (4. Under such transformations we have V(φ) → V(φ) h(x) . ˜ ˜ [hα . One starts from scalar ﬁelds which take their values in the homogeneous space.48) . Subsequently. which now depends on n ﬁelds. (4. i j ˜ fαβγ hγ + ξα ξβ Rij (H) .46) V −1 ∂µ V = Qµ + Pµ .

−→ h−1 (x) Pµ (x) h(x) . When this is not done. whose ﬁeld equations are algebraic and are solved by (4. Pν ] . so that (4. Pν] ] = 0 . which presupposes (4.26). The result can be written as. After substituting the result one obtains the second-order form.50) (4. After imposing the gauge condition on V(x) one obtains the coset representative V(φ(x)).54) Clearly this Lagrangian is invariant under the group G. Furthermore both Pµ and Qµ are invariant under rigid G-transformations.5) There are several ways to write down the Lagrangian of the corresponding nonlinear sigma model. L = − 1 tr Pµ P µ .53) L = 1 tr Dµ V −1 Dµ V .49) Hence Qµ acts as a gauge ﬁeld associated with the local H transformations.and in a second-order form. the theory is invariant under Grigid × Hlocal with both groups acting linearly. de Wit: Supergravity 67 By allowing ourselves to perform such local gauge transformations. Qµ (x) Pµ (x) −→ h−1 (x) Qµ (x) h(x) + h−1 (x) ∂µ h(x) .51) from the beginning. (4. At this stage one still has the full gauge invariance with respect to local H transformations and one can impose a gauge restricting V to a coset representative.51). Qν ] = −[Pµ . (4. the . but until that point V will just be an unrestricted spacetime dependent element of the group G.52) Here we made use of the commutation relations (4.47) reads V −1 Dµ V = Pµ . In the ﬁrst one regards the gauge ﬁeld Qµ as an independent ﬁeld. Hence we write (4. as soon as one imposes a gauge and restricts V to a coset representative parametrized by certain ﬁelds φi . = ∂[µ Pν] + [Q[µ . Dµ V = ∂µ V − VQµ . Eventually we will ﬁx this gauge freedom.51) Just as before. Obviously the Lagrangian must be invariant under both the rigid G transformations and the local H transformations. Fµν (Q) D[µ Pν] = ∂µ Qν − ∂ν Qµ + [Qµ . (4. we introduced new degrees of freedom into V associated with the group H. It is convenient to introduce a corresponding H-covariant derivative. However. 2 (4. From (4. one derives the Cartan-Maurer equations (4.48) we derive the following local H-transformations. 2 One can interpret this result in a ﬁrst.B.

The corresponding spaces have dimension 70 and 42. As we intend to demonstrate. Also here we can proceed in ﬁrst. Matter ﬁelds are assigned to a representation of the local H group. The exceptional groups are noncompact and are divided by their maximal compact subgroups. In supergravity we are often dealing with sigma models based on homogeneous. whereas the gauge-ﬁxed approach leads to unsurmountable diﬃculties (at least. this strategy allows for a systematic approach. Later in this chapter we will be dealing with the E7(7) /SU (8) and E6(6) /U Sp(8) coset spaces. .or secondorder formalism. These target spaces are usually noncompact and Riemannian. respectively. so that H is the maximally compact subgroup of G. symmetric target spaces. gauge ﬁelds have to transform under the rigid group G. In this form the local H invariance is absent. 2 (4. the matter ﬁelds transform nonlinearly under the group G with transformations that take the form of φ-dependent H-transformations.54) leads to the standard form of the nonlinear sigma model. We emphasize that the presentation that we followed so far was rather general. the strategy is to postpone this gauge choice till the end. This combined subgroup still generates a representation of the group G. In this form it is easy to see how to couple matter ﬁelds to the sigma model in a way that the invariance under the isometries remains unaﬀected.36). Therefore.31). for the spaces of interest). In order to deal with complicated supergravity theories that involve homogeneous spaces. In ﬁrst-order form the equation (4. so that they couple to the sigma model ﬁelds through the connection Qµ that appears in the covariant derivatives. Upon choosing a gauge. However. so that one is always dealing with a manifest linearly realized symmetry group Grigid × Hlocal . the maximally symmetric spaces that we considered served only as an example. Usually the ﬁelds will remain invariant under the group G as long as one does not ﬁx the gauge and choose a speciﬁc coset representative. Gauge Theory and Strings residual subgroup is such that the H transformations are linked to the G transformations and depend on the ﬁelds φi . but it is now realized in a nonlinear fashion. L = − 1 gij (φ) ∂µ φi ∂ µ φj .68 Unity from Duality: Gravity. determined by (4.55) where the target space metric is given by (4. but the invariance under G is still there and realized as target space isometries generated by corresponding Killing vectors. It is straightforward to demonstrate that (4. gauge ﬁelds cannot couple in this way as their gauge invariance would be in conﬂict with the local invariance under the group H.47) will acquire some extra terms that depend on the matter ﬁelds.

for sigma models based on homogeneous target spaces one can proceed in a way in which all transformations remain linearly realized. G/H (4. The consistency of this procedure is obvious as (4. where we have introduced a coupling constant g to keep track of the new terms introduced by the gauging.4 Gauged nonlinear sigma models 69 Given a nonlinear sigma model with certain isometries. one can gauge some or all of these isometries in the usual way: one elevates the parameters of the isometry group (of a subgroup thereof) to arbitrary functions of the spacetime coordinates and introduces the necessary gauge ﬁelds (with their standard gauge-invariant Lagrangian containing a kinetic term) and corresponding covariant derivatives. the expressions for Qµ and Pµ will change. Hence the covariant derivative of V is now changed by the addition of the corresponding (dynamical) gauge ﬁelds Aµ which take their values in the corresponding Lie algebra (which is a subalgebra of the Lie algebra associated with G). as we will discuss in the next section. the original rigid invariance under G transformations from the left is now broken by the embedding of the new gauge group into G. To adopt this approach is extremely important for the construction of gauged supergravity theories.58) When imposing a gauge condition.36) (in the gauge where V equals the coset representative). H . Qν ] . Of course.57) Because Qµ and Pµ now depend on the gauge connections Aµ . We will always use the second-order formalism so that the H-connection Qµ will not be an independent ﬁeld. Hence. (4. Pν ] − g V −1 Fµν (A)V 1 − 2 g V −1 Fµν (A)V .51). (4. according to Qµ = Q(0) − g[V −1 Aµ V]H . µ (0) Pµ = Pµ − g[V −1 Aµ V]G/H . The modiﬁcations caused by the new minimal couplings are minor and the eﬀects can be concisely summarized by the Cartan-Maurer equations. but the derivative is now covariantized and modiﬁed by the terms depending on the new gauge ﬁelds Aµ .56) Dµ V(x) = ∂µ V(x) − V(x) Qµ (x) − g Aµ (x) V(x) . Fµν (Q) = = D[µ Pν] = ∂µ Qν − ∂ν Qµ + [Qµ . As explained above. They remain expressed by (4.51) is fully covariant. Since these new gauge transformations involve the isometry group they must act on the group element V as a subgroup of G.B. [Pµ . the last result for Pµ exhibits precisely the Killing vectors (4. With this change. When gauging isometries in a generic nonlinear sigma model . de Wit: Supergravity 4.

The latter may give rise to groundstates with nonzero cosmological constant. They contain an extra parameter g. as before). 5 Gauged maximal supergravity in 4 and 5 dimensions The maximally extended supergravity theories introduced in Section 3 were obtained by dimensional reduction from 11-dimensional supergravity on a hypertorus. and arises when compactifying seven coordinates of D = 11 supergravity on the sphere S 7 . in some dimensions the G-invariance is not realized at the level of the action. in 4 dimensions the 28 Abelian vector ﬁelds do not constitute a representation of the group E7(7) .70 Unity from Duality: Gravity. Supersymmetry requires the presence of extra terms of order g and g 2 in the Lagrangian. which is the non-Abelian coupling constant. In this compactiﬁcation the four-index ﬁeld strength acquires a nonzero values when all its indices are in the four-dimensional spacetime. The modiﬁcations in the matter sector arise through the order g contributions to Qµ . In these theories the scalar ﬁelds parametrize a G/H coset space (cf. Generically the ﬁelds transform as follows. . where for simplicity we assumed a single isometry. which has an SO(8) isometry group. Gauge Theory and Strings (cf. rather than on the vector ﬁelds. one replaces the derivatives according to ∂µ φ → ∂µ φi −Aµ ξ i (φ). The group emerges as the gauge group of the compactiﬁed theory formulated in 4 dimensions. An obvious gauging in D = 4 dimensions is based on the group SO(8).55)). The graviton is invariant. In the next section we will discuss the application of this formulation to gauged supergravity. Note that Pµ and Qµ are invariant under the new gauge group (but transform under local H-transformations. as the Lagrangian has a manifest SO(8) invariance and there are precisely 28 vector ﬁelds [63]. Table 12) and the group G is also realized as a symmetry of the full theory. For example. This gauging has an obvious KaluzaKlein origin. This turns out to be possible and the corresponding theories are known as gauged supergravities. we concentrate on the maximal gauged supergravities in D = 4 and 5 spacetime dimensions. (4. However. In this case the group G is realized by electric-magnetic duality and acts on the ﬁeld strengths. To explain the construction of gauged supergravity theories. In this theory the E7(7) invariance group is broken to a local SO(8) group so that the resulting theory is invariant under SU (8)local × SO(8)local . We return to electric-magnetic duality in Section 5. the (Abelian) gauge ﬁelds transform linearly under G and the fermions transform linearly under the group H. Apart from the gauge ﬁeld interactions there are fermionic masslike terms of order g and a scalar potential of order g 2 .3. It is an obvious question whether these theories allow an extension in which the Abelian gauge ﬁelds are promoted to non-Abelian ones. but at the level of the combined ﬁeld equations and Bianchi identities.

Both these exceptional Lie groups can be introduced in terms of 56-dimensional matrices22 . This implies that 15 of the 27 gauge ﬁelds become associated with the non-Abelian group SO(6). Note that the tensors Fµν . it turns out that many other subgroups of E7(7) can be gauged. In this solution the ﬁve-index (self-dual) ﬁeld strength acquires a nonzero value whenever the ﬁve indices take all values in either S 5 or in the ﬁve-dimensional spacetime. In D = 5 dimensions they are E6(6) and U Sp(8).B. where SO(6) is realized as a local gauge group. Representation assignments for the various supergravity ﬁelds with respect to the groups G and H. de Wit: Supergravity 71 Table 16. The solutions is to convert these 12 gauge ﬁelds to antisymmetric tensor ﬁelds. In D = 4 dimensions these groups are E7(7) and SU (8). Type-IIB supergravity has a manifest SL(2) invariance and the isometry group of S 5 is SO(6). In D = 5 dimensions the possible gaugings are not immediately clear. so that the symmetry group of the Lagrangian equals the SL(2) × SO(6) subgroup of E6(6) . 22 Strictly speaking the isotropy groups are SU (8)/Z2 and U Sp(8)/Z2 . Before continuing with supergravity we ﬁrst discuss some basic features of the two coset spaces E7(7) /SU (8) and E6(6) /U Sp(8). The Lagrangian can thus be written in a form that is invariant under U Sp(8)local × SO(6)local × SL(2) [29]. respectively. . Gµν and/or Hµν denote the ﬁeld strengths of the vector ﬁelds and/or (for D = 5) possible tensor ﬁelds. which leaves 12 Abelian gauge ﬁelds which are charged with respect to the same group. v ijIJ 28 + 28 56 27 + 27 27 + 27 SU (8) E7(7) U Sp(8) E6(6) 1 1 1 1 8 1 8 1 However. Also in 5 dimensions other gauge groups are possible. We will brieﬂy comment on this issue at the end of the section. as the Abelian gauge transformations of these 12 ﬁelds will be in conﬂict with their transformations under the SO(6) gauge group. Gµν . Again the Kaluza-Klein scenario can serve as a guide. This poses an obvious problem. a eµ i ψµ Fµν . as there is no obvious 27-dimensional gauge group. While D = 11 supergravity has no obvious compactiﬁcation to ﬁve dimensions. type-IIB supergravity has a compactiﬁcation on the sphere S 5 . Hµν 1 56 1 27 χijk 56 1 48 1 uijIJ .

IJ IJ (z1 . ΛIJKL z KL + ΣIJKL zKL (5. the dimension of E7(7) equals 63+70 = 133. R) in a pseudoreal basis. which has dimension 63. Consider now inﬁnitesimal transformations of the form. (5. (5. R) are E∗ = ω E ω . ΣIJKL = [K ΛI J = −ΛJ I . (5. . δzIJ δz IJ = = ΛIJKL zKL + ΣIJKL z KL . The deﬁning properties of elements E of Sp(56.4) E −1 = Ω E † Ω . where ω and Ω are given by ω= 0 1 1 0 . The matrices ΛIJ KL are associated with its maximal compact subgroup. acting on a pseudoreal vector (zIJ . Gauge Theory and Strings 5. Hence the (zIJ . which is equal to U (28).5) is invariant. (5.1 On E7(7) /SU(8) and E6(6) /USp(8) cosets We discuss the E7(7) and E6(6) on a par for reasons that will become obvious. K. J. for which the ΣIJKL is fully antisymmetric and the generators are further restricted according to ΛIJKL = δ[I ΛJ]L] . .1) where ΛIJKL and ΣIJKL are subject to the conditions (ΛIJKL )∗ = ΛIJKL = −ΛKLIJ .72 Unity from Duality: Gravity. Let us now consider the E7(7) subgroup. the number of the noncompact generators minus the number of compact ones . . (ΣIJKL )∗ = ΣKLIJ . z KL ) span a 56-dimensional vector space. To deﬁne the groups we consider the fundamental representation. 8.6) Obviously the matrices ΛIJ generate the group SU (8). ΛI I = 0 . 1 24 εIJKLMN P Q ΣMN P Q . In passing we note that the real subgroup (in this pseudoreal representation) is equal to the group GL(28). This group is the group of electric-magnetic dualities of maximal supergravity in D = 4 dimensions. (5. where the indices are antisymmetrized index pairs [IJ] and [KL] and I. Because SU (8) is the maximal compact subgroup.2) The corresponding group elements constitute the group Sp(56. since ΣIJKL comprise 70 real parameters. z KL ) with z IJ = (zIJ )∗ . . L = 1. Ω= 1 0 0 −1 . z2 ) = z1 z2IJ − z1IJ z2 .3) The above properties ensure that the sequilinear form.

rather than to SU (8). These numbers conﬁrm that we are indeed dealing with E6(6) and its maximal compact subgroup U Sp(8). However. ΛIJKL = δ[I ΛJ]L] .7) IJKLMN P Q εIJKLMN P Q z IJ KL MN P Q z z z +ε zIJ zKL zMN zP Q . It is straightforward to show that these matrices close under commutation and generate the group E7(7) .9) ΣIJKL = ΩIM ΩJN ΩKP ΩLQ ΣMN P Q . however. To see this we observe that the following restrictions are preserved by the group. (5. while the diﬀerence between the numbers of noncompact and compact generators equals 42 − 36 = 6. (ΩIJ )∗ = ΩIJ . J ΩIK ΩKJ = −δI . (5.9) the 56-dimensional representation deﬁned by (5. Altogether we thus have 36 + 42 = 78 generators. [K ΛI J = −ΛJ I . for which the restrictions are rather similar. For further information the reader is encouraged to read the appendices of [53]. The maximal compact subgroup U Sp(8) thus has dimension 64 − 28 = 36. E7(7) has another maximal 63-dimensional subgroup. which is not compact. de Wit: Supergravity 73 is equal to 70 − 63 = 7. This is the group U Sp(8). deﬁnes the group SL(8. satisfying ΩIJ = −ΩJI . ΩIJ ΣIJKL = 0 .8) Now we restrict ourselves to the subgroup of U (8) that leaves ΩIJ invariant. J4 (z) = zIJ z JK zKL z LI − 1 (zIJ z IJ )2 4 1 + 96 (5. Another subgroup is the group E6(6) . The group E7(7) has a quartic invariant. zIJ = ±ΩIK ΩJL z KL . one of them is that the contraction ΣIKLM ΣJKLM is traceless. plays no role in the following.B. To show this one needs a variety of identities for selfdual tensors [64]. ΩIJ z IJ = 0 .1) and where the diagonal blocks correspond to the group SL(8). while there are 70 − 28 = 42 generators associated with ΣIJKL . .6) with ΛIJ and ΣIJKL real. This is the group SL(8). The other restrictions on the generators concern ΣIJKL . It is possible to choose conventions in which the E7(7) matrices have a diﬀerent block decomposition than (5. Altogether we have the conditions. which. Because of the constraints (5. (5.1) is reducible and decomposes into two singlets and a 27 and a 27 representation.10) The ﬁrst one suppresses the singlet representation and the second one projects out the 27 or the 27 representation. Here one introduces a skew-symmetric tensor ΩIJ . Λ[I K ΩJ]K = 0 . We note that the subgroup generated by (5. R).

etc.12) where the rank-4 antisymmetric tensor Σ satisﬁes the algebraic restrictions appropriate for the exceptional group. Table 12. This is the group SL(6) × SL(2).11) which plays a role in the E6(6) invariant Chern-Simons term in the supergravity Lagrangian. Because the spinless ﬁelds always appear in nonpolynomial form. or 5 dimensions. as well as in the gauging. which may diﬀer from the notations used in the original references where one sometimes uses left cosets. and which plays a role in many of the known gaugings. Hence.74 Unity from Duality: Gravity. be able to bring out the main features of the G/H structure. 0 (5. the 56-bein is reducible for E6(6) . As explained in the previous section. but we will use the reducible version in order to discuss the two theories on a par. so that SL(2) remains as a rigid invariance group of the Lagrangian. (5. For those and other details we refer to the original literature [29. 5 where the coset structure plays an important role. that will be relevant in the following. hopefully. There is another maximal subgroup of E6(6) . Our notation will be based on a description in terms of right cosets. V(x) = exp 0 Σ(x) Σ(x) . just as in the previous sections.1).2 On ungauged maximal supergravity Lagrangians An important feature of pure extended supergravity theories is that the spinless ﬁelds take their values in a homogeneous target space (cf. One starts by introducing a so-called 56-bein V(x). without getting entangled in issues that depend on the spacetime dimension. which has dimension 35 + 3 = 38. which is noncompact. we assume that the 56-bein transforms under the exceptional group from the left and under the local SU (8) (or . Schematically. We will be rather cavalier about numerical factors. In this way we will. depending on whether we are in D = 4. where we have listed these spaces). 63]. We will not be complete here but sketch a number of features of the maximal supergravity theories in D = 4. spinor conventions. where the gauge group is embedded into the group SL(6). deﬁned by J3 (z) = z IJ z KL z MN ΩJK ΩLM ΩN I . which is a 56 × 56 matrix that belongs to the group E7(7) or E6(6) . Gauge Theory and Strings The group E6(6) has a cubic invariant. A coset representative is obtained by exponentiation of the generators deﬁned in (5. it is vital to exploit the coset structure explained in the previous section in the construction of the supersymmetric action and transformation rules. 5.

where left cosets were chosen. uijKL ΩIK ΩJL = Ωik Ωjl uklIJ . we have ﬁxed the gauge with respect to local SU (8) or U Sp(8). for E7(7) . uijIJ uijKL − vijIJ v ijKL uijIJ vijKL − vijIJ uijKL = = IJ δKL .12) is a coset representative. The latter are the indices associated with the local SU (8) or U Sp(8). uijIJ uklIJ − v ijIJ vklIJ uijIJ v klIJ − v ijIJ uklIJ or. In the above the row indices are ([IJ].13) ijKL KL −v (x) ukl (x) with the usual conventions uijIJ = (uijIJ )∗ and vijIJ = (v ijIJ )∗ .13) gauge ﬁxing is not assumed.3) the inverse V −1 can be expressed in terms of the complex conjugates of the submatrices of V. ij = δkl . Observe also that (5. i.. 8 (5. According to (5. [KL]). The notation of the submatrices is chosen such as to make contact with [63].15) (5. V(x) = (5. except that the antisymmetrized Kronecker symbols on the right-hand sides are replaced according to ij ij δkl → δkl − 1 Ωkl Ωij .e. whereas in (5. = 0. (5. In this case the (pseudoreal) matrices uij IJ ± ΩIK ΩJL v ijKL and their complex conjugates deﬁne (irreducible) .B.14) V −1 (x) = klIJ kl v (x) u KL (x) Consequently we derive the identities. Observe that the indices of the matrix are antisymmetrized index pairs [IJ] and [ij]. conversely. IJ uij (x) vijKL (x) .g. (5. (5. 0. and the column indices are ([ij].16) The corresponding equations for E6(6) are identical. de Wit: Supergravity 75 U Sp(8)) from the right. e. [kl]). uijIJ ΩIJ = 0 .18) with similar identities for the v ijIJ . 8 IJ IJ δKL → δKL − 1 ΩKL ΩIJ . The 56-bein can be decomposed in block form according to ij −vklIJ (x) u IJ (x) . upon interchanging V and V −1 .17) Furthermore the matrices u and v vanish when contracted with the invariant tensor Ω and they are pseudoreal.

(5. Gauge Theory and Strings elements of E6(6) corresponding to the 27 and 27 representations. Subsequently we evaluate the quantities Qµ and Pµ . v ijIJ ∂µ uklIJ − uijIJ ∂µ v klIJ . 8 (5. Qµ ijmn klmn Pµ Pµ ijpq Qµklpq . Qµ ijkl ijkl Pµ = = uijIJ ∂µ uklIJ − vijIJ ∂µ v klIJ . We note the identity uij IJ + ΩIK ΩJL v ijKL ukl IJ − ΩIM ΩJN vklMN ij = δkl − 1 Ωkl Ωij .25) Because of the underlying Lie algebra the connections Qµ ij satisfy Qµij = −Qµj i and Qµi i = 0.22) or.76 Unity from Duality: Gravity. Hence Qµ ijkl must satisfy the decomposition.20) (5. with the reality condition. to the 42-dimensional representation of U Sp(8). . Qµ ijkl = δ[i Qµ j]l] . (5.24) uikIJ ∂µ ujkIJ − vikIJ ∂µ v jkIJ . Therefore it transforms according to the 70-dimensional representation of SU (8).19) In this case we can thus decompose the 56-bein in terms of a 27-bein and a 27-bein. ijkl Pµ = 1 24 εijklmnpq Pµ mnpq . respectively. so that Qµ ij equals Qµ ij = 2 3 [k (5. as well as an extra symmetry condition in the case of U Sp(8) (cf. with the reality condition. (5.23) Pµ = Ωim Ωjn Ωkp Ωlq Pµ mnpq . Likewise Qµ transforms as a connection associated with SU (8) or U Sp(8).21) ijkl The important observation is that Qµijkl and Pµ are subject to the same constraints as the generators of the exceptional group listed in the previijkl ous section. V −1 ∂µ V = which leads to the expressions. Hence. (5. ijkl (5. Pµ is fully antisymmetric and subject to a reality constraint.9)).

27).26) Observe that the use of the Lie algebra decomposition for G/H is crucial in deriving these equations. (5. δQµ ij ijkl δPµ = = 2 3 jklm Σjklm Pµ iklm − Σiklm Pµ . The ﬁrst one is the observation that any variation of the 56-bein can be written. Fµν (Q)i j ijkl D[µ Pν] 77 = ∂µ Qν ij − ∂ν Qµ ij + Q[µ ik Qν]kj = − 4 P[µjklm Pν]iklm . up to a local H-transformation. as is also obvious from the decomposition (5.28) is consistent with both groups G and H. (5. this establishes that the SU (8) tensors Qµ and Pµ can be assigned to the adjoint representation of the group G. (5. up to complex conjugation and possible contractions with H-covariant tensors.29) Observe that. where Σ is an H-covariant expression proportional to the supersymmetry parameter i and the fermion ﬁelds χijk . Such decompositions are an important tool for dealing with the spinless ﬁelds in this nonlinear setting. Furthermore Σ must satisfy the restrictions associated with the exceptional group. Under this variation the quantities Qµ and Pµ transform systematically. Fixing the gauge prematurely and converting to a speciﬁc coset representative for G/H would lead to unsurmountable diﬃculties.. Continuing along similar lines we turn to a number of other features that are of interest for the Lagrangian and transformation rules. the generators denoted by k in the previous section). .9). As was stressed above. Before ﬁxing a gauge. In particular this applies to supersymmetry transformations. 3 ijkl = ∂[µ Pν] + 2Q[µ m[i Pν] jkl]m = 0.27). δuijIJ = −Σijkl v klIJ . δvijIJ = −Σijkl uklIJ . i. the variation (5. we can avoid the nonlinearities completely and carry out the calculations in a transparent way.B.6) or (5. as δV = V or. up to a local H-transformation.e. (5.20).28) 0 Σ Σ 0 . Dµ Σijkl = ∂µ Σijkl + 2Qµ m[i Σjkl]m . in terms of submatrices. any variation of V can be decomposed into (5.e. Because Σ takes the form of an H-covariant tensor. (5. The supersymmetry variation can be written in the form (5. de Wit: Supergravity Furthermore we can evaluate the Maurer-Cartan equations (4.52).27) where Σijkl is the rank-four antisymmetric tensor corresponding to the generators associated with G/H (i. Hence it must be of the form Σijkl ∝ ¯[i χjkl] .

28) applied in diﬀerent orders on the 56-bein. we have [δ1 . δ2 ] V ∝ (¯i γ µ 1 2i [i jkl]m − ¯i γ µ 2 1 i) V 0 Pµ Pµ 0 . 1 and 2 . (5. Here we distinguish the Einstein-Hilbert term for gravity. For the ﬁrst term ijkl we note that δ1 Σ2 leads to a term proportional to Pµ .32) This is precisely a spacetime diﬀeomorphism. Let us now turn to the action. Gauge Theory and Strings The supersymmetry variation of the spinor χijk contains the quantity ijkl Pµ .2 system. so that up to proportionality constants we must have a variation. The last term represents the 1 Noether coupling term for the spin-0/spin. this contribution can be written in the form [δ1 .33) This Lagrangian is manifestly invariant with respect to E7(7) or E6(6) . respectively. ω) − 1 ψµ γ µνρ (∂ν − 1 ων γab )δi + 1 Qν ij ψρj 2 2 2 ab l 1 − 12 χijk γ µ (∂µ − 1 ωµ γab )δk + 3 Qµ kl χijl − ¯ 2 2 ijkl 1 − 12 χijk γ ν γ µ ψνl Pµ .9). (5.30) The veriﬁcation of the supersymmetry algebra on V is rather easy. Hence up to a number of ﬁelddependent H-transformations. e−1 L1 = j ab 1 ¯i − 2 R(e. ¯ ijkl 1 96 Pµ µ Pijkl (5. combined with two supersymmetry parameters. the Dirac Lagrangian and the nonlinear sigma model associated with the G/H target space. the terms in the Lagrangian pertaining to the graviton. For D = 4 the fermion ﬁelds are chiral spinors and we have to add the contributions from the . which incorporates the spacetime derivatives of the spinless ﬁelds. 0 Σ2 Σ2 0 . (5. gravitini. of the form (¯1 γ µ 2 m ) Pµ . as follows from (5.78 Unity from Duality: Gravity. spinors and scalars take the following form. up to a local H-transformation proportional to Qµ . ijkl δχijk ∝ Pµ γ µ l . the Rarita-Schwinger Lagrangian for the gravitini. the supersymmetry commutator closes on V into a spacetime diﬀeomorphism (up to terms of higher-order in the spinors that we suppressed).20).6) or (5. Apart from higher-order spinor terms.31) The last term is just an inﬁnitesimal H-transformations. Under two consecutive (ﬁeld-dependent) variations (5. Taking into account the various H-covariant combinations in the actual expressions implied by (5. δ2 ] V = V 0 2 δ[1 Σ2] 2 δ[1 Σ2] 0 +V 0 Σ1 Σ1 0 .

However. Here we disregard such details and concentrate on the symmetry issues. ψµ and χijk . the combined equations of motion and the Bianchi identities are invariant under the group E7(7) . Hence we deﬁne a U Sp(8) covariant ﬁeld strength for the vector ﬁelds. In D = 5 dimensions the vector ﬁelds Bµ transform as the 27 representation of E6(6) . µν (5. up to terms quartic in the fermion ﬁelds. for D = 5 we are dealing with so-called symplectic Majorana spinors.11) and a moment coupling with the µν fermions. equal to IJ IJ Fij = (uijIJ − v ijKL ΩIK ΩJL )(∂µ Bν − ∂ν Bµ ) . the phenomenon is not restricted to . uijIJ +v ijKL ΩIK ΩJL . It is impossible to construct an invariant action just for the vector ﬁelds and one has to make use of the scalars.35) gives the full supergravity Lagrangian invariant under rigid E6(6) and local U Sp(8) transformations. can lead to an equivalent set of ﬁeld equations and Bianchi identities.3 Electric–magnetic duality and E7(7) For D = 4 the Lagrangian is not invariant under E7(7) but under a smaller group. Any additional dependence would aﬀect the invariance under E6(6) . and which can be used to convert E6(6) to U Sp(8) indices. where the symmetry group of ﬁeld equations and Bianchi identities can be bigger than that of the Lagrangian.35) where we distinguish the kinetic term. We continue the discussion of the D = 4 theory in the next section. which are diﬀerent for spaceIJ time dimensions D = 4 and 5. Observe that the dependence on the spinless ﬁelds is completely implicit. and the Lagrangian is manifestly invariant under the corresponding transformations. and where diﬀerent Lagrangians not related by local ﬁeld redeﬁnitions. The vector ﬁelds bring in new features. However.33) and (5. µν (5.34) The invariant Lagrangian of the vector ﬁelds then reads.B. which can be written in terms of the 27-beine. de Wit: Supergravity 79 spinors of opposite chirality. This situation is typical for D = 4 theories with Abelian vector ﬁelds. a Chern-Simons interaction associated with the E6(6) cubic invariant (5. Here Oij denotes a covariant tensor antisymmetric in both spacei time and U Sp(8) indices and quadratic in the fermion ﬁelds. as this requires to ﬁrst introduce the concept of electric-magnetic duality. The result obtained by combining the Lagrangians (5. L2 = 1 − 16 e Fij Fµνkl Ωik Ωjl µν IJ KL MN 1 − 12 εµνρσλ Bµ ∂ν Bρ ∂σ Bλ ΩJK ΩLM ΩN I µν 1 + 4 e Fij Oij . so that they satisfy the reality KL constraint Bµ IJ = ΩIK ΩJL Bµ . which acts on the vector ﬁelds (but not necessarily on the 56-bein) according to a 28-dimensional subgroup of GL(28). 5.

36) ±α Together Fµν and G± α comprise the electric and magnetic ﬁelds and inµν ductions. Its simplest form arises in Maxwell theory in four-dimensional (ﬂat or curved) Minkowski space. constant. so that the rotation is associated with a 56 × 56 matrix.37) which are obviously invariant under real. .80 Unity from Duality: Gravity. Gauge Theory and Strings 4 dimensions and can occur for antisymmetric tensor gauge ﬁelds in any even number of spacetime dimensions (see. [52]). .) are decomposed into selfdual and ±α anti-selfdual components Fµν (which. These ﬁeld strengths (in the case at hand we have 28 of them. rotations of the ﬁeld strengths F ± and G± ±α Fµν G± β µν −→ U W Z V ±α Fµν G± β µν . n × n submatrices and β n denotes the number of independent gauge potentials.g. which commute with the Lorentz group and rotate the electric and magnetic ﬁelds and inductions according to E ↔ H and B ↔ D.36). This condition amounts to an integrability condition. which can only have a solution (for generic Lagrangians) provided that the matrix .38) where U α . but for the moment we will remain more general and label the ﬁeld strengths by α.g. see. The relevant question is whether the rotated equations (5. are related by complex conjugation) and so are the ﬁeld strengths G± α that appear in µν the ﬁeld equations. real. labelled by antisymmetric index pairs [IJ]. Wαβ and Z αβ are constant. ﬁeld strengths. in Minkowski space. The Bianchi identities and equations of motion for the Abelian gauge ﬁelds take the form ∂µ F + − F − α µν = ∂ µ G+ − G− µν α = 0. This duality can be generalized to any D = 4 dimensional ﬁeld theory with Abelian vector ﬁelds and no charged ﬁelds. [51]). In N = 8 supergravity we have 56 such ﬁeld strengths of each duality. More precisely. β. which are deﬁned by G± α = ± µν ∂L 4i · e ∂F ± αµν (5. does there exist a new Lagrangian depending on the new. e. rotated. Vαβ . The 4-dimensional version has been known for a long time and is commonly referred to as electric-magnetic duality (for a recent review of this duality in supergravity. (5. e. such that the new tensors Gµν follow from this Lagrangian as in (5. (5.37) can again follow from a Lagrangian. . where one can perform (Hodge) duality rotations. so that the gauge ﬁelds enter the Lagrangian only through their (Abelian) ﬁeld strengths.

Furthermore. which are related to the index pairs [IJ] of the 56-bein (cf.B.39) We distinguish two subgroups of Sp(2n. Namely. R)23 . (5. Furthermore the Lagrangian is not uniquely deﬁned (it can always be reparametrized via an electric-magnetic duality transformation) and neither is its invariance group. Given the fact that we can rotate the ﬁeld strengths by electric-magnetic duality transformations. one can show that in Minkowski spaces of dimensions D = 4k. which usually requires the other ﬁelds in the Lagrangian to transform as well. a generic theory does not have such an invariance group. Z TV = V TZ . de Wit: Supergravity 81 is an element of the group Sp(2n. One is the invariance group of the combined ﬁeld equations and Bianchi identities. we assign diﬀerent indices to the ﬁeld strengths and the underlying gauge groups than to the 56-bein V. this group equals SO(n. . we label the ﬁelds strengths by independent index pairs [AB]. (5. to remain in the context of the pseudoreal basis used previously. there exist diﬀerent Lagrangians with diﬀerent symmetry groups. this invariance group is not necessarily realized as a symmetry of the Lagrangian. but maximal supergravity is known to have an E7(7) ⊂ Sp(56. R) was ﬁrst derived in [65]. These issues are extremely important when gauging a subgroup of the invariance group. we form the linear combinations. whose Bianchi identities and equations of motion are the same (modulo a linear transformation) and are invariant under the same group (which contains the symmetry groups of the various Lagrangians as subgroups).40) 2µν µν 2 23 Without any further assumptions. For rank-k antisymmetric gauge ﬁelds in D = 2k + 2 dimensions. In this respect our presentation is more in the spirit of a later treatment in [66] for N = 2 vector multiplets coupled to supergravity (duality invariances for these theories were introduced in [67]). but in the context of a duality invariance rather than of a reparametrization. Of course. U TW = W TU . R).13)) in a way that we will discuss below. More precisely. Observe that these groups do not constitute an invariance of the theory. 1µν µν 2 +AB F+AB = 1 (i G+ AB − Fµν ) . as this requires the gauge group to be contained in the invariance group of the Lagrangian. R). R). is usually smaller and restricted by Z = 0 and U −1 = V T . R) invariance. the rotations of the ﬁeld equations and Bianchi identities associated with n rank-(k − 1) antisymmetric gauge ﬁelds that are described by a Lagrangian. +AB F+ AB = 1 (i G+ AB + Fµν ) . but merely characterize an equivalence class of Lagrangians. The fact that the symplectic redeﬁnitions of the ﬁeld strengths constitute the group Sp(2n. the subgroup associated with the matrices U equals GL(n). However. constitute the group Sp(2n. n. The subgroup that is a symmetry of the Lagrangian. This implies that the submatrices satisfy the constraint U TV − W TZ = V U T − W Z T = 1 . (5.

For convenience. [63]). The various Lagrangians are thus encoded in Sp(56. (5. R) matrices E. unless it belongs to a GL(28) subgroup which can act on the gauge ﬁelds themselves. (5. Gauge Theory and Strings Anti-selfdual ﬁeld strengths (F−AB .e. we redeﬁne the 56-bein by absorbing the . F2 ).82 Unity from Duality: Gravity. the real GL(28) subgroup is induced by corresponding linear transformations of the vector ﬁelds. E= UIJAB VKLAB VIJCD UKLCD . the 56-bein plays a crucial role. it can be absorbed into the 56-bein V.42) one can straightforwardly determine the relevant terms in the Lagrangian. R) according to the matrices E speciﬁed in (5. In the basis (5. To exhibit how one can deal with a continuous variety of Lagrangians. in elements of E7(7) \Sp(56. On the other hand.3). From (5. Hence the matrix E allows the ﬁeld strengths and the 56-bein to transform under E7(7) in an equivalent but nonidentical way. F− AB ) follow from complex conjuga1µν 2µν tion.42) is unique (cf. The tensor F+ ij is an SU (8) covariant ﬁeld strength that µν appears in the supersymmetry transformation rules of the spinors. R) matrix E (so that it satisﬁes the conditions (5. On this basis the ﬁeld strengths rotate under Sp(56.3)). when E ∈ E7(7) . R)/GL(28). In order to establish this consistency. One could consider absorbing this matrix into the deﬁnition of the ﬁeld strengths (F1 . but such a redeﬁnition cannot be carried out at the level of the Lagrangian. symmetric.41) On the basis of E7(7) and SU (8) covariance. which is simply deﬁned by the above condition.3) the generators of GL(28) have a block decomposition with SO(28) generators in both diagonal blocks and identical real. let us AB remember that the tensors Fµν and Gµν AB are related by (5. up to multiplication by GL(28) from the right and multiplication by E7(7) from the left. The relation involves a constant Sp(56. which appears in the moment couplings in the Lagrangian.36) and this relationship must be consistent with E7(7) . the relation among the ﬁeld strengths must have the form. which are manifestly invariant under diﬀerent subgroups of E7(7) . V −1 E F+ AB 1µν F+AB 2µν = F+ ij µν +kl Oµν . i. Without going into the details we mention that the chirality + and duality of Oµν is severely restricted so that the structure of (5. 28 × 28 matrices in the oﬀ-diagonal blocks.42) + where Oµν is an SU (8) covariant tensor quadratic in the fermion ﬁelds and independent of the scalar ﬁelds.

+ The SU (8) covariant ﬁeld strength Fµνij will appear in the supersymmetry transformation rules for the fermions. .1).. they must also satisfy (5.6). +AB L3 = − 1 e Fµν F +CD µν [(u + v)−1 ]ABij (uijCD − v ijCD ) 8 +AB − 1 e Fµν [(u + v)−1 ]ABij O+µν ij 2 + h. . satisfying Im ΣABCD + ΛABCD = 0 . . the matrix multiplying the two ﬁeld strengths in (5. . µνij (5.39). . this subgroup is generated on V by matrices Λ and Σ.44) ν µ and which take the form. . ˆ V(x) = E−1 V(x) = −v ij CD 83 uijAB (x) (x) −v kl AB (x) uklCD (x) .. but only after a proper conversion of the I. rather than with I.45) AB where the 28 × 28 matrices satisfy [(u + v)−1 ]ABij (uijCD + v ijCD ) = δCD .38) and (5. J. R). however. (5. the imaginary parts of 24 Such symmetry properties follow from the symmetry under interchanging index pairs in the products (uijAB − vijAB ) (uklAB + vklAB ) and (uijAB + vijAB ) (uijCD + vijCD ). . indices. Because the matrix E−1 V is an element of Sp(56. . (5. According to the ˆ analysis leading to (5. AB Fµν = ∂µ AAB − ∂ν AAB . . . but with indices A. to A. Note.e. one has to make a certain choice for the matrix E. J. . (5. B. that the E7(7) tensors Qµ and Pµ are not aﬀected by ˆ the matrix E and have identical expressions in terms of V and V..B. B.46) Clearly the Lagrangian depends on the matrix E. and is equal to +AB +ij Fµν = (uijAB + v ijAB ) F+ − (uijAB + vijAB ) Oµν . de Wit: Supergravity matrix E. This is not the case for the terms in the Lagrangian that contain the Abelian ﬁeld strengths. The gauge ﬁelds transform under the real subgroup (i. .43) ˆ where we have to remember that V is now no longer a group element of E7(7) . In order that the Lagrangian be invariant under a certain subgroup of E7(7) .45) is symmetric under the interchange of [AB] ↔ [CD]24 . . just as in (5.47) In order to be a subgroup of E7(7) as well. .c. (5.

CD TEF . When the latter are charged under the gauge group. TCD ] = fAB. 5. is E6(6) × SO(1. in D = 5 this obstruction can be avoided. Gauge Theory and Strings the generators act exclusively on the 56-bein). In D = 4 dimensions. . 6−p) involves 15 non-Abelian gauge ﬁelds and 12 antisymmetric tensor ﬁelds. 1). the gauging of SO(p. then there is a potential obstruction to the gauging as the gauge invariance of these gauge ﬁelds cannot coexist with the nonAbelian gauge transformations. after assigning the various ﬁelds to representations of the gauge group embedded in E7(7) or E6(6) . the Lagrangian is manifestly symmetric under E6(6) . Typically this conversion of vector into tensor ﬁelds leads to terms that are inversely proportional to the gauge coupling [68]. However. we will mainly concentrate on this theory. Therefore. Hence the Abelian ﬁeld strengths are changed to non-Abelian ones and derivatives of the scalars are covariantized according to ∂µ V → ∂µ V − gAAB TAB V . because (charged) vector ﬁelds can alternatively be described as antisymmetric rank-2 tensor ﬁelds. one must start from a Lagrangian that is symmetric under the desired gauge group. as one can deduce from the symmetry groups that are realized in maximal supergravity in higher dimensions.48) where the gauge group generators TAB are 56 × 56 matrices which span a subalgebra of maximal dimension equal to the number of vector ﬁelds. the T-tensor The gauging of supergravity is eﬀected by switching on the gauge coupling constant. which requires one to make a suitable choice of the matrix E. This aspect is most nontrivial for the D = 4 theory. When eﬀecting the gauging the vector ﬁelds may decompose into those associated with the non-Abelian gauge group and a number of remaining gauge ﬁelds. In D = 5 dimensions. The biggest group whose existence can be inferred in this way. Only the gauge ﬁelds themselves and the spinless ﬁelds can transform under this gauge group. embedded in the Lie algebra of E7(7) or E6(6) . µ (5.49) It turns out that the viability for a gauging depends sensitively on the choice of the gauge group and its corresponding embedding. A large variety of symmetry groups exists. The structure constants of the gauge group are given by EF [TAB . For instance. (5. because they are not realized as tensor gauge ﬁelds. The latter can transform under the gauge group.4 Gauging maximal supergravity. so this subtlety does not arise. However.84 Unity from Duality: Gravity. in view of electric-magnetic duality. which is the group that one obtains from the D = 5 Lagrangian upon reduction to D = 4 dimensions.

(5. 3 AB ijkl − 1 g Fµν PAB . ijkl PAB = = QAB ijkl 1 ijklmnpq PAB mnpq 24 ε [k δ[i QAB j]l] .51) where V −1 TAB V = QAB ijmn klmn PAB These modiﬁcations are the result of the implicit dependence of Qµ and Pµ on the vector potentials AAB .48) of the Cartan-Maurer equations. This modiﬁcation was already noted in (4. QAB ij ijkl PAB = = 2 3 uikIJ (∆AB ujkIJ ) − vikIJ (∆AB v jkIJ ) . on which the E7(7) transformations act in the basis that is appropriate for the ﬁeld strengths.53) v ijIJ (∆AB uklIJ ) − uijIJ (∆AB v klIJ ) where ∆AB u and ∆AB v indicate the change of submatrices in V induced by multiplication with the generator TAB . (5. in the basis appropriate for V. (5. The leading variations are induced by the modiﬁcation (5.B. The order-g terms in the Maurer-Cartan equation yield ¯ . 2 (5. . Note that we could have expressed ˆ the above quantities in terms of the modiﬁed 56-bein V. For convenience we now restrict ourselves to D = 4 dimensions. Introducing the gauging leads directly to a loss of supersymmetry. The fact that the matrices TAB generate a µ subalgebra of the algebra associated with E7(7) . implies that the quantities QAB and PAB satisfy the constraints. (5. provided we change the generators TAB into ˆ TAB = E−1 TAB E .52) while QABij is antihermitean and traceless. It is straightforward to write down the explicit expressions for QAB and PAB .57) and takes the form Fµν (Q)i j ijkl D[µ Pν] = = AB − 4 P[µjklm Pν]iklm − g Fµν QAB i j .50) PAB ijpq QABklpq . because the new terms in the Lagrangian lead to new variations.54) This is done below. de Wit: Supergravity 85 to write down a corresponding Lagrangian requires an even number of tensor ﬁelds. When establishing supersymmetry of the action one needs the CartanMaurer equations at an early stage to cancel variations from the gravitino kinetic terms and the Noether term (the term in the Lagrangian proportional to χψµ Pν ).

We will concentrate on properties which are perhaps less obvious. ˆ ˆ ˆ ˆ where ∆AB u and ∆AB v are the submatrices of TAB V.58) This shows that the SU (8) covariant T -tensors can be assigned to a representation of E7(7) . as its variations yield again the same tensor. δTijkl mn δTijkl = = kl Σjmnp Timnp − kl 1 jmnpqrst Σimnp Tqrst + 24 ε pmn pqrs 4 1 − 24 εijklpqrs Σmntu Ttu . δL = ρ µν i 1 4 g(¯j γ γ ψρ 1 + 288 εijklmnpq − ¯i γ ρ γ µν ψρj ) QAB i j (uklAB + v klAB ) F+ µνkl χijk γ µν l PAB mnpq (ursAB + v rsAB ) F+ ¯ µνrs (5. Gauge Theory and Strings the leading variations of the Lagrangian. Some of them are rather obvious (such as Tiijk = 0). and follow rather straightforwardly from the deﬁnition. They are linearly proportional to the fermion ﬁelds and read.55) + h. which allows one to derive. Tijkl = = j 3 4 QAB i 1 2 (uklAB + v klAB ) (5. ˆ The T -tensor is thus a cubic product of the 56-bein V which depends in a nontrivial way on the embedding of the gauge group into E7(7) . First we observe that SU (8) covariantized variations of the T -tensor are again proportional to the T -tensor.86 Unity from Duality: Gravity. which is a special feature of D = 4 dimensions. these properties are generic. The ﬁrst variation is proportional to an SU (8) tensor Tijkl . which is known as the T -tensor. It satisﬁes a number of important properties. This includes the SU (8) covariant derivative of the T -tensor. Apart from the ˆ distinction between V and V.57) mnAB ˆ vijCD (∆AB uklCD ) ).56) ˆ ˆ uimCD (∆AB ujmCD ) − vimCD (∆AB v jmCD ) (uklAB + v klAB ) . Before completing the analysis leading to a consistent gauging we stress that all variations are from now on expressed in terms of the T -tensor. (5. This property will play an important role below. Along the same lines as before we can show that the SU (8) tensors QAB and PAB transform according to the adjoint representation of E7(7) . Another component of the T -tensor appears in the second variation and is equal to mn Tijkl = = 1 2 PAB ijkl 1 2 (umnAB + v mnAB ) − ˆ uijCD (∆AB vklCD ) (umnAB +v (5.28). 3 Σp[ijk Tl] j Σklmn Timn . which follows directly from (5.58) upon the . These variations are induced by (5.27) and (5.c.

A1 . To see how this can be the case. The three SU (8) covariant tensors. let us analyze the SU (8) content of the T -tensor. ¯ 3 (5. A2ijkl and Aijk. de Wit: Supergravity 87 substitutions δ → Dµ and Σ → Pµ . √ ¯i δg ψµ = − 2g Aij ¯j γµ . (5. the T -tensor is cubic in the 56-bein.lmn χijk χlmn + h.lmn have certain symmetry prop1 3 erties dictated by the way they appear in the Lagrangian (5.58). A viable gauging requires that the T -tensor satisﬁes a number of rather nontrivial identities. First of all. √ j ¯ ¯i Lmasslike = g e 1 2 A1ij ψ i γ µν ψν + 1 Ajkl ψµ γ µ χjkl µ 2 6 2i +Aijk.60) Finally at order g 2 one needs a potential for the spinless ﬁelds. as we will discuss shortly. The transformation properties were given in (5. . A3: 28 + 420 + 1176 + 1512 . P (V) = g 2 jkl 2 1 24 |A2i | − 1 |Aij |2 · 1 3 (5.c. and as such is constitutes a certain tensor that transforms under E7(7) . irrespective of the gauge group. to cancel the variations (5. because they were introduced for the purpose of cancelling the variations (5.B. where we made use of the fact that the T -tensor consists of a product of the fundamental times the adjoint representation of E7(7) . This implies that these tensors transform under SU (8) according to the representations A1: 36 .55). A2: 28 + 420 . which depend only on the spinless ﬁelds.59). but the new terms in the Lagrangian and transformation rules have a universal form.55) we need masslike terms in the Lagrangian.61) These last three formulae will always apply. Let us ﬁrst describe these new terms. 1 δg χijk = −2g A2lijk ¯l . Hence the T -tensor comprises the representations.62) . irrespective of the gauge group.59) whose presence necessitates corresponding modiﬁcations of the supersymmetry transformations of the fermion ﬁelds. must be linearly related to the T -tensor. (5. To be speciﬁc. Note that the tensors Aij . A2 is fully antisymmetric in [jkl] and A3 is antisymmetric in [ijk] as well as in [lmn] and symmetric under the interchange [ijk] ↔ [lmn]. A1 is symmetric in (ij). 56 × 133 = 56 + 912 + 6480 . As we mentioned already. A2 and A3 .

which gave rise to Tijkl mn Tijkl = = = 3 − 4 A2ijkl + 3 A1 δi . Ajkl = − 4 Ti 2i 3 [jkl] .58) the T -tensor must be restricted by suppressing complete representations of E7(7) in order that its variations and derivatives remain consistent. One then ﬁnds that there is yet another constraint. 912 = 36 + 36 + 420 + 420 . Indeed this was found by explicit calculation.88 Unity from Duality: Gravity. This implies that the T -tensor is decomposable into A1 and A2 .65) Observe that a contraction with the ﬁrst upper index is also equal to zero.66) Although we concentrated on the D = 4 theory. we note that there is a mismatch between (5.lmn 3 (5. respectively. Ti [jk]i = 0.63) and (5. as follows from the deﬁnition (5. we should stress once more that many of the above features are generic and apply in other dimensions. whereas A3 is not an independent tensor and can be expressed in terms of A2 . For instance. which suppresses the 28 representation of the T -tensor. (5.63) 6480 = 28 + 28 + 420 + 420 + 1280 + 1280 + 1512 + 1512 . (5. (5. Hence the T -tensor transforms under E7(7) according to the 912 representation which decomposes into the 36 and 420 representations of SU (8) and their complex conjugates residing in the tensors A1 and A2 . with the result 56 = 28 + 28 . Gauge Theory and Strings The representations on the right-hand side can be decomposed under the action of SU (8). Comparing these representations to the SU (8) representations to which the tensors A1 −A3 (and their complex conjugates) belong. In view of (5. [m n] j[k l] Aijk. √ n] 1 − 108 2 εijkpqr[lm Tpqr . but not suﬃcient as one also needs nontrivial identities quadratic in the T -tensors in order to deal with the variations of the Lagrangian of order g 2 . This proves that the T tensor cannot contain the entire 6480 representation of E7(7) . the unrestricted T -tensors in D = 5 and 3 dimensions belong .62).64) Note that these conditions are necessary. 2 4 − 3 δ[i Tjkl] .56). Aij = 1 ikj 4 21 Tk . so that it must consist of the 28 + 36 + 420 representation of SU (8) (and its complex conjugate).

for the 5-dimensional theory they were also realized in [29] and in [72]. We established the existence of two supersymmetric gravitational backgrounds. These terms lose their topological nature when gauging some of the E8(8) isometries [69]. starting with the Einstein-Hilbert Lagrangian for gravity and the Rarita-Schwinger Lagrangians for the gravitino ﬁelds.67) In these cases a successful gauging requires the T -tensor to be restricted to the 351 and the 1 + 3875 representations. which decompose as follows under the action of U Sp(8) and SO(16). see [74. 6 Supersymmetry in anti-de Sitter space In Section 3. 89 = 3: 248 × 248 = 1 + 248 + 3875 + 27 000 + 30 380 . In D = 3 dimensions there is no guidance from Kaluza-Klein compactiﬁcations and one has to rely on the group-theoretical analysis described above. gave rise to the gauge groups SO(8). respectively. (5. 1 + 3875 = 1 + 135 + 1820 + 1920 .1 we presented the ﬁrst steps in the construction of a generic supergravity theory. for D = 5 A3 is again dependent while for D = 3 there is an independent tensor A3 associated with the 1820 representation of SO(16). namely ﬂat Minkowski space and anti-de Sitter space with a cosmological 25 The D = 3 theory has initially no vector ﬁelds. Gaugings can also be constructed via a so-called Scherk-Schwarz reduction from higher dimensions [73].68) These representations correspond to the tensors A1 and A2 . but those can be included by adding Chern-Simons terms. (5. 351 = 36 + 315 . SO(5) and SO(6). For explorations based on the group-theoretical analysis explained above. respectively. the ﬁrst gaugings were to some extent motivated by corresponding KaluzaKlein compactiﬁcations. respectively25 D D = 5: 27 × 78 = 27 + 351 + 1728 . 75]. de Wit: Supergravity to the following representations of E6(6) and E8(8) . . We close with a few comments regarding the various gauge groups that have been considered. In that case there exists a large variety of gauge groups of rather high dimension [69]. To give a really exhaustive classiﬁcation remains cumbersome. Noncompact gauge groups were initiated in [71] for the 4-dimensional theory. The S 7 and the S 4 [70] compactiﬁcations of 11dimensional supergravity and the S 5 compactiﬁcation of IIB supergravity. As we mentioned at the beginning of this section.B.

respectively. possibly augmented with the R-symmetry generators e associated with rotations of the supercharges. +. For D > 7 the superalgebra is no longer simple [3]. In D = 3 spacetime dimensions the anti-de Sitter group SO(2. In that case the bosonic subalgebra coincides with the anti-de Sitter algebra. the hypersurface is invariant under linear transformations that leave the metric ηAB = diag (−. . and HR = U Sp(2N ) for D = 6. Gauge Theory and Strings constant proportional to g 2 . σ σ Rµνρσ = g 2 (gµρ δν − gνρ δµ ) .1 was incomplete and in general one will need to introduce additional ﬁelds.5. this hypersurface is deﬁned by −(Y − )2 − (Y 0 )2 + Y 2 = ηAB Y A Y B = −g −2 .14)). Q} anticommutator. 2)/SO(D−1.90 Unity from Duality: Gravity. 2). 2. As we discussed in Section 2. 2) is not simple. In this section we will mainly be dealing with simple anti-de Sitter supersymmetry and we will always assume that 3 < D ≤ 7. Anti-de Sitter space is isomorphic to SO(D−1.1) Because both ﬂat Minkowski space and anti-de Sitter space are maximally symmetric. These transformations constitute the group SO(D − 1. These N supercharges transform under a compact R-symmetry group. (6.2) Obviously. each transforming as a spinor under the anti-de Sitter group. In contrast to this. . Denoting the extra coordinate of the embedding space by Y − . whose generators will ap¯ pear in the {Q. but one needs extra bosonic generators that transform as high-rank antisymmetric tensors under the Lorentz group (see also [76]). whose bosonic generators correspond to the Poincar´ group. 0. D − 1. . there exists an (N -extended) super-Poincar´ algebra associated with ﬂat e Minkowski space of any dimension. so that we have coordinates Y A with A = −. . There exist N -extended versions where one introduces N supercharges. . (6. . 1. The 1 D(D + 1) generators 2 . they have 1 D(D + 1) independent isometries which comprise 2 the Poincar´ group or the group SO(D − 1. . where g was some real coupling constant proportional to the the inverse anti-de Sitter radius. 7. +) invariant. According to (4. +. the R-symmetry group is in general not the same as in Minkowski space. 1) and thus belongs to the coset spaces that were discussed extensively in Section 4. The algebra of e the combined bosonic and fermionic symmetries is called the anti-de Sitter superalgebra. HR = U (N ) for D = 5. The two cases are clearly related and ﬂat space is obtained in the limit g → 0. −. its bosonic subalgebra can no longer be restricted to the sum of the anti-de Sitter algebra and the R-symmetry algebra.10) it is possible to describe anti-de Sitter space as a hypersurface in a (D + 1)-dimensional embedding space. . we have HR = SO(N ) for D = 4. (3. as can for instance be seen from the expression of the Riemann curvature (cf. 2). Note again that the derivation in Section 3. according to Table 9.

2) group in the spinor representation.5) Here the matrices ΓAB . Qβ } = ¯ [MAB .4) contains the (anti-)commutation relations ¯ {Qα . These curves can be avoided by unwrapping S 1 . de Wit: Supergravity denoted by MAB act on the embedding coordinates by MAB = YA ∂ ∂ − YB . Γb } = 2 η ab 1. which will be deﬁned later. As we alluded to earlier. keeping the remaining generators MAB corresponding to the Lorentz subalgebra unchanged. It is now easy to evaluate the commutation relations for the MAB [MAB . Here one rescales the generators according to o u M−A → g −1 PA . +. ∂Y B ∂Y A 91 (6. Spatial inﬁnity is a timelike surface which cannot be reached by timelike geodesics. . B = a.B = a (6. . Our gamma matrices satisfy the Cliﬀord property {Γa . A = −. (6. which has the topology of RD . In the limit g → 0. . which in addition to (6. b = 0. are the generators of SO(D − 1. . (6. For later use we record the (simple) anti-de Sitter superalgebra. There are many ways to coordinatize anti-de Sitter space. . . the anti-de Sitter algebra can be realized by the following combination of gamma matrices Γa in D-dimensional Minkowski space. Qα ] = 1 − 2 (ΓAB )αβ M AB . MAB = 1 ΓAB = 2 1 2 Γab 1 2 Γa for for A. e On spinors. this algebra changes its form when considering N supercharges which rotate under R-symmetry. There exist no Cauchy surfaces in this space.3) where we lower and raise indices by contracting with ηAB and its inverse η AB . so that one ﬁnds the universal covering space denoted by CadS.6) with a. where η ab = diag (−. b . The relation with the Minkowski case proceeds by means of a so-called Wigner-In¨n¨ contraction. because the R-symmetry generators will appear on the ¯ right-hand side of the {Q.4) Anti-de Sitter space has the topology of S 1 [time] × RD−1 [space] and has closed timelike curves. +) is the D-dimensional . MCD ] = ηBC MAD − ηAC MBD − ηBD MAC + ηAD MBC .B. . 1 ¯ 2 (Q ΓAB )α . . but we will avoid using explicit coordinates. D−1. 78]. Q → g −1/2 Q. 1. the generators PA will form a commuting subalgebra and the full algebra contracts to the super-Poincar´ algebra. Q} anticommutator. Any attempt to determine the outcome of some evolution or wave equation from a spacelike surface requires fresh information coming from a timelike inﬁnity which takes a ﬁnite amount of time to arrive [77.

Of central importance is the quadratic Casimir operator of the isometry group SO(D − 1. This result can be proven for any symmetric. e.92 Unity from Duality: Gravity.9) = Dµ ( 1 ¯2 Γa 1 ) + 1 g (¯2 Γab 1 ) eµb . for a complete theory there are additional contributions. Using the transformation rules (3. The appropriate formulae were given at the end of Section 4. 2 (6.5.2 (cf. we obtain for the vielbein. deﬁned by C2 = − 1 M AB MAB .8) vanishes and the proof is elementary (see. 2 2 The ﬁrst term corresponds to a spacetime diﬀeomorphism and the second one to a tangent space (local Lorentz) transformation. [δ1 . the reader is advised to consult Section 2. homogeneous. Let us now brieﬂy return to the supersymmetry algebra as it is realized on the vielbein ﬁeld.11) The commutator of two supersymmetry transformations yields an inﬁnitesimal generalcoordinate transformation and a tangent-space Lorentz transformation.1. For example. 2) has more Casimir operators when D > 3. (6. Gauge Theory and Strings Lorentz-invariant metric26 . but nevertheless the above terms remain and 26 Note that when the gravitino is a Majorana spinor. [80]). Concerning the R-symmetry group in anti-de Sitter space.44) and (4. for example [79]). We remind the reader that ﬁelds in anti-de Sitter space are multi-component functions of the anti-de Sitter coordinates that rotate irreducibly under the action of the Lorentz group SO(D − 1. For scalar ﬁelds. . but these are of higher order in the generators and will not play a role in the following. 2). To make contact between the masslike terms in the wave equations and the properties of the irreducible representations of the anti-de Sitter group. 1). which involves the appropriately covariantized D’Alembertian 2adS . it is important that we establish the relation between the wave operator for ﬁelds that live in anti-de Sitter space.45)) and from them one can derive. and the quadratic Casimir operator C2 . the quantities Γ AB should satisfy the same Majorana constraint. Here we consider only the gravitational sector of the theory. (4.g. the second term in (6. δ2 ] eµ a = 1 2 ¯2 Γa δ1 ψµ − 1 ¯1 Γa δ2 ψµ 2 (6. C2 = 2adS g=1 Lorentz + C2 .8) Lorentz where C2 is the quadratic Casimir operator for the representation of the Lorentz group to which the ﬁelds have been assigned. space (see. which we will discuss in Section 6.7) The group SO(D − 1.

it satisﬁes (4.9) should vanish when 1 and 2 are Killing spinors.e.B.. which is a Killing spinor in the anti-de Sitter space. (6. / (6.2 for generic coset spaces.. Further clariﬁcation from an algebraic viewpoint will be given later in Section 6.10). Indeed. We have already discussed the interpretation of masslike terms for the gravitino. (x) must satisfy the Killing spinor equation (3. so that P 2 is a Casimir operator. In the anti-de Sitter background. A scalar chiral supermultiplet in 4 spacetime dimensions consists of a scalar ﬁeld A. 2 because Dµ (g ¯2 Γνρ 1 ) = −g 2 gµ[ρ ξν] is antisymmetric in µ and ν. For supermultiplets in anti-de Sitter space this is no longer the case. Therefore both the gravitino and the vielbein are left invariant under supersymmetry. a pseudoscalar ﬁeld B and a Majorana spinor ﬁeld ψ.15). In the notation of this section.3. In anti-de Sitter space the supersymmetry transformations of the ﬁelds are proportional to a spinor parameter (x). were the gravitino ﬁeld vanishes.9) should be realized uniformly on all the ﬁelds.9) is a compensating target space transformation of the type we have been discussing extensively in Section 4. δA δψ = 1 ¯ψ .15) so that the gravitino ﬁeld remains zero under supersymmetry.1 Anti-de Sitter supersymmetry and masslike terms In ﬂat Minkowski space all ﬁelds belonging to a supermultiplet are subject to ﬁeld equations with the same mass. As for all Killing vectors. the diﬀeomorphism with parameter ξ µ = 1 ¯2 Γµ 1 . so that the combination of symmetries on the right-hand side of (6.2) .1) where we made the anti-de Sitter vierbein and spin connection explicit. following (3. Indeed. this equation reads a ∂µ − 1 ω ab γab + 1 g eµ γa 4 2 = 0. is an anti-de Sitter Killing vector (i. δB = 1 i¯γ5 ψ . This phenomenon will be now illustrated below in a speciﬁc example. so that masslike terms will not necessarily be the same for diﬀerent ﬁelds belonging to the same multiplet. We allow for two constants a and b in the supersymmetry transformations. the parameters of the supersymmetry transformations are Killing spinors satisfying (3. i. higher derivatives can be decomposed into the Killing vector and its ﬁrst derivative.38)). because the momentum operators commute with the supersymmetry charges. we ﬁnd Dµ (g ¯2 Γνρ 1 ) = −g 2 gµ[ρ ξν] in the case at hand. 4 4 = ∂ (A + iγ5 B) − (a A + ib γ5 B) . de Wit: Supergravity 93 (6. 6. The last term in (6. which we parametrize as follows. namely a scalar chiral supermultiplet in D = 4 spacetime dimensions. The Killing vector can be decomposed into the 1 D(D + 1) 2 Killing vectors of the anti-de Sitter space.e.

the supersymmetry transformation of ψ equals δψ = D(A + iγ5 B) − m(A + iγ5 B) − g(A − iγ5 B) . The second term in D (6. we consider the variation of the fermionic ﬁeld equation. After some Fierz reordering we obtain the result.6) The last equation follows from the Dirac equation. Again we have to take into account that derivatives on the supersymmetry parameter are not equal to zero. (6.9) but the correct Lorentz transformation is only reproduced provided that a − b = 2g. / The commutator yields the Riemann curvature of the anti-de Sitter space. Obviously.3) We point out that derivatives acting on (x) occur in this calculation at an intermediate stage and should not be suppressed in view of (6. which is proportional to g. possibly up to a ﬁeld equation. [2adS + 3g 2 − m2 ] ψ = 0. we consider the closure of the supersymmetry algebra on the spinor ﬁeld.1). e.5). b = m−g. so that the last term is just 2 the Dirac equation with mass m. can be understood as a compensating S-supersymmetry transformation associated with auxiliary ﬁelds in the supergravity sector (see. δ2 ] yields the correct coordinate transformation ξ µ Dµ on the ﬁelds A and B. Now we note that the right-hand side should constitute a coordinate transformation and a Lorentz transformation. To determine the constants a and b and the ﬁeld equations of the chiral multiplet. [2adS + 2g 2 − m(m + g)] B = 0 . [81]). the coordinate transformation coincides with (6. 1 [δ1 .5) and the fermionic ﬁeld equation equals (/ + m)ψ = 0. (6. This yields the following second-order diﬀerential equations [2adS + 2g 2 − m(m − g)] A = 0 . Dψ 2 2 (6.g. can be accounted for by introducing an auxiliary ﬁeld to the supermultiplet. The coeﬃcient of the ﬁrst term in δψ has been chosen such as to ensure that [δ1 . we ﬁnd a = m+g. δ2 ]ψ = ξ µ Dµ ψ + 16 (a− b) ¯2 γ ab 1 γab ψ − 1 ξ ρ γρ [/ + 1 (a+ b)ψ] . The third term. In an anti-de Sitter space of arbitrary dimension D this equation then reads [2adS + 1 D(D − 1)g 2 − m2 ]ψ = 0 . If we now deﬁne m = 1 (a + b). they produce terms proportional to g which turn out to cancel in the above commutator. Namely. which is proportional to m. 4 (6. / (6. D D 2 D.94 Unity from Duality: Gravity. Gauge Theory and Strings In this expression the anti-de Sitter vierbein ﬁeld has been used to contract the gamma matrix with the derivative. which gives rise to the wave operator 2adS + 1 [/ D]− m2 .7) .4) Consequently. However. In order to construct the corresponding ﬁeld equations for A and B. one evaluates (/ − m)(/ + m)ψ.

In the same way as before. (6. 2+ 1D−2 R = 2 + 1 D(D − 2) g 2 .8) Dµ (∂µ Aν − ∂ν Aµ ) − m2 Aν = 0 . νn ) (n + 1) Dµ ∂[µ Cν1 ···νn ] − m2 Cν1 ···νn = 0 . This will be elucidated later in Section 6. we do reproduce the g 2 dependence in the ﬁrst two equations (6. for D = 4. whose ﬁeld equation reads (antisymmetrizing over indices ν1 .10) The g 2 term in the ﬁeld equations for the scalar ﬁelds can be understood from the observation that the scalar D’Alembertian (in an arbitrary gravitational background) can be extended to a conformally invariant operator (see. 4 4D−1 (6.11) (6.8) reads D2 Aν −[Dµ . 1 D(D − 1).6). for D = 4 agrees with the last equation of (6. otherwise one can impose this equation as a gauge condition.6). 6. This leads to27 Dµ Aµ = 0. so that (6. . . vector and tensor ﬁelds.2. The quadratic Casimir operator of the Lorentz group takes the values 0. D − 1 and n(D − n) for scalar. [81]). e. Combining this result with (6. .12) which seems the obvious candidate for a massless wave operator for scalar ﬁelds.9) and 27 When m = 0. de Wit: Supergravity 95 which. Observe that the Dirac operator D is also conformally / invariant and so is the wave equation associated with the Maxwell ﬁeld. Consequently.B. the role of mass is quite diﬀerent in anti-de Sitter space as compared to ﬂat Minkowski space. Dν ]Aµ −m2 Aν = 0 or. in spite of the fact that they belong to the same supermultiplet [82]. Using (6. this leads to [2adS + n(D − n)g 2 − m2 ] Cν1 ···νn = 0 . 6.9) This can be generalized to an antisymmetric tensor of rank n.8) we can now determine the values for the quadratic Casimir operator for the representations described by scalar. spinor. Indeed. (6.g. spinor.6) all have diﬀerent mass terms. .12. . A striking feature of the above result is that the ﬁeld equations (6. (6. For future applications we also evaluate the Proca equation for a massive vector ﬁeld. vector and 8 tensor ﬁelds respectively. in anti-de Sitter space [2adS + (D − 1)g 2 − m2 ] Aµ = 0 .7.

scalar C2 spinor C2 vector C2 tensor C2 = − 1 D(D − 2) + m2 . In a number of cases this decoupling is more extreme and one obtains a so-called singleton representation which does not have a smooth Poincar´ limit. 6. spinor. a rank-n and a rank-(D − n − 2) tensor gauge ﬁeld are equivalent on shell (also in curved space).2 the bound on C2 implies that m2 ≥ 0. which ensures the stability of an anti-de Sitter background against small ﬂuctua1 tions of the scalar ﬁelds. as that is given by the value for C2 . which uniquely identiﬁes the massless representations and their values of C2 . This can also be seen from the observation that (massless) antisymmetric tensor gauge ﬁelds of rank n = D − 2 are on-shell equivalent to massless scalar ﬁelds.13). m2 is not the coeﬃcient in the mass term of the Kleinscalar Gordon equation. In those cases there is no decoupling of a representation e that could be identiﬁed as massless and therefore there remains a certain ambiguity in the deﬁnition of “massless” representations. For massless representations of higher spin there is a decoupling of degrees of freedom. (6. We thus derive speciﬁc values for C2 for massless scalar. (6. While we concluded above that these tensor ﬁelds lead to C2 = 0. (6.8) yields the following values for the quadratic Casimir operators. respectively. For spin.14) This result is known as the the Breitenlohner-Freedman bound [82]. = m2 . In the next section we will derive a rather general lower bound on the value of C2 for the lowest-weight representations of the anti-de Sitter algebra (cf. vector and tensor ﬁelds upon putting m = 0. 1 µ2 ≥ − 4 (D − 1)2 . Hence. which implies that the masslike terms for scalar ﬁelds can have a negative coeﬃcient µ2 subject to the inequality. In the next section we study unitary representations of the anti-de Sitter algebra and this study will conﬁrm some of the results found above.13). while.10. massless 1 scalar ﬁelds have C2 = − 4 D(D − 2) according to (6. we see that the interpretation of the mass parameter is not straightforward in the context of anti-de Sitter space. for spinor 2 and vector ﬁelds. The diﬀerence may not be entirely surprising in view of the fact that the antisymmetric tensor Lagrangian is not conformally invariant for arbitrary values of D.26)). 4 = − 1 D(D − 1) + m2 . 1 2 whereas for spin-0 we ﬁnd that m ≥ − 4 . m and m do correspond to the mass terms in the Dirac and Proca equations. with m2 as deﬁned in (6. for m2 = 0.96 Unity from Duality: Gravity. 8 = m2 .13) For spinless ﬁelds. Gauge Theory and Strings (6. contrary . The fact that the value of C2 does not depend on the rank of the tensor ﬁeld is in accord with the fact that.

(6. 2). is noncompact. At the end of the next section. 84. The anti-de Sitter isometry group. are quantized in integer units in order to have single-valued functions. associated with motions along the circle. R) × SL(2. These 2 generators. . refer to space indices.16) Obviously the generators of the spatial rotations are the operators Mab with a. At any rate. ± Ma = −iM0a ± M−a . . have diﬀerent hermiticity properties. the generator M−0 is related to the energy operator when the radius of the anti-de Sitter space is taken to inﬁnity. The eigenvalues of this generator. however. 2 First.15) Here we note that the cover group of SO(D − 1. The connection between the local ﬁeld theory description and the anti-de Sitter representations tends to be subtle. (6. 2) ∼ (SL(2.B.2 Unitary representations of the anti-de Sitter algebra In this section we discuss unitary representations of the anti-de Sitter algebra. 2) has the generators 1 Γµν 2 and 1 Γµ . For these representations the generators are anti-hermitean. Note that we have changed notation: here and henceforth in this section the indices a. . The remaining generators M−a and M0a are combined into D −1 pairs of mutually conjugate operators. unless one passes to the covering space CadS. The compact subgroup of the anti-de Sitter group is SO(2) × SO(D − 1) corresponding to rotations of the compact anti-de Sitter time and spatial rotations. acting on spinors which are ﬁnite-dimensional objects. SO(D − 1. = . b = 1. We refer to [83] for some of the original work. The energy operator H will thus be deﬁned as H = −iM−0 . we brieﬂy return to the issue of massless representations. we have established the existence of two diﬀerent ﬁeld representations that describe massless. .17) 28 The case of D = 3 is special because SO(2. In order to underline the general features we will stay as much as possible in general spacetime dimensions D > 328 . It is convenient to decompose the 1 D(D+1) generators as follows. b. de Wit: Supergravity 97 to the scalar ﬁeld Lagrangian. where we discuss unitary representations of the anti-de Sitter algebra. † MAB = −MAB . D − 1. spinless states which correspond to diﬀerent values for the anti-de Sitter Casimir operator C2 . which implies that unitary representations will be inﬁnitely dimensional. . 85] where part of this work was reviewed. and to [80. R))/Z2 . 6. . . (6.

Ma } − 1 (Mab )2 2 2 + − = H(H − D + 1) + J 2 − Ma Ma . C2 = − 1 M AB MAB 2 + − = H 2 − 1 {Ma .20) In simple cases. (s+ . so that irreducible representations are characterized by two spin values. (6. let us already brieﬂy indicate how some of the other (anti-)commutators of the simple anti-de Sitter superalgebra decompose (cf. A restricted class of representations will be discussed in a sequel. Their dimension is equal to (2s+ + 1)(2s− + 1) and 2 2 2 J 2 = 2(J+ + J− ) with J± = s± (s± + 1). its value is well known. −2(H δab + Mab ) . a more general discussion of all possible representations requires a more technical set-up and is outside the scope of these lectures. (6. deﬁned by 1 J 2 = − 2 (Mab )2 .21) The SO(D−1) representations for D > 5 are speciﬁed by giving the eigenvalues of additional (higher-order) SO(D − 1) Casimir operators. Summarizing: J2 = s(s + 1) for D = 4 . Ma ] = ± ± [Ma . (6. 0.98 Unity from Duality: Gravity. (6. The anti-de Sitter commutation relations then read ± [H.19) where J 2 is the total spin operator: the quadratic Casimir operator of the rotation group SO(D − 1). the Ma play the role of raising and lowering operators: when ± applied to an eigenstate of H with eigenvalue E. For D = 4 it is expressed in terms of the “spin” s which is an integer for bosons and a half-integer for fermions and the spin-s representation has dimension 2s + 1 and J 2 = s(s + 1). but in passing. Mb ] ± ±Ma . s− ). (6. Mb ] = − + [Ma .5)) {Qα . In this section we restrict ourselves to the bosonic case. the corresponding rotation group SO(4) is the product of two SU (2) groups. 2 . application of Ma yields a state with eigenvalue E ± 1. Q† } β = H δαβ − 1 iMab (Γa Γb Γ0 )αβ 2 + − + 1 (Ma Γa (1 + iΓ0 ) + Ma Γa (1 − iΓ0 ))αβ .18) = ± Obviously. 2s+ (s+ + 1) + 2s− (s− + 1) for D = 5 . We also give the Casimir operator in this basis. Gauge Theory and Strings + − satisfying (Ma )† = Ma . For D = 5.

J . 2 ± [Ma . For unitary lowest-weight representations one can derive various lower bounds.24) The unitarity upper bound (6. all separated by integers. We now turn to irreducible representations of the anti-de Sitter algebra (6. C2 = E0 (E0 − D + 1) + J 2 . J = 0 . This will be discussed later in Section 6. This is therefore not a continuous representation but a so-called lowest-weight representation. as we shall see below. (6. we derive at once the eigenvalue of the quadratic Casimir operator associated with this representation in terms of E0 and J 2 .25) Since C2 is a Casimir operator. For a unitary representation the + − Ma Ma term in (6.19) is positive. so that in principle one could restrict oneself to fermionic operators only and employ the projections (1±iΓ0)Q as the basic lowering and raising operators. 2 this number will depend on whether the eigenvalues E take integer or halfinteger values. J and satisﬁes − Ma |E0 . However.22) For the anti-de Sitter superalgebra.18).18) applied to the ground state |E0 . this value holds for any state belonging to the corresponding irreducible representation. The ground state of this representation (which itself transforms as an irreducible representation of the rotation group and may thus be degenerate) is denoted by |E0 . Continuous representations cover the whole range of eigenvalues E extending from −∞ to ∞. Qα ] = ∓ 1 (Γa (1 ∓ iΓ0 ) Q)α . 2 99 (6. de Wit: Supergravity [H . all the bosonic operators can be expressed as bilinears of the supercharges. Substituting the condition (6. (6. which implies that the Casimir operator is bounded by C2 ≤ − 1 (D − 1)2 + J 2 + E − 1 (D − 1) 4 2 2 minimal . Qα ] = − 1 i(Γ0 Q)α . (6. when there is a state with some eigen− value E0 that is annihilated by all the Ma .23) where the subscript indicates that one must choose the minimal value that J 2 + (E − 1 (D − 1))2 takes in the representation.B.24) in the expression (6.3. Among other things. then only states with eigenvalues E > E0 will appear in the representation. Because application of Ma leads to the states with higher and lower eigenvalues E. We start with the observation that the energy operator can be diagonalized so that we can label the states according to their eigenvalue ± E. For real values of E0 the Casimir .23) on C2 is primarily useful for continuous representations. we expect the representation to cover an inﬁnite range of eigenvalues.

1. for scalar ﬁelds (J 2 = 0) this is just the Breitenlohner-Freedman bound [82]. They generally lead to a lower bound for E0 and thus to a corresponding lower bound for C2 .26) As we already discussed at the end of the previous section. Gauge Theory and Strings E 6 E0 + 5 E0 + 4 E0 + 3 E0 + 2 E0 + 1 E0 s . Additional restrictions based on unitarity will be derived shortly. permissible values for E0 with the same value for C2 . States of the spinless representation in terms of the energy eigenvalues E and the angular momentum j.j 0 1 2 3 4 5 6 s s s s s s s s s s s Fig. operator is bounded by C2 ≥ J 2 − 1 (D − 1)2 . symmetric tensors Y a1 ···al of rank l = j.26) there can exist a degeneracy in the sense that there are two possible. These two values correspond to two diﬀerent solutions of the ﬁeld equations subject to diﬀerent boundary conditions at spatial inﬁnity. Alternatively we can e construct highest-weight representations.100 Unity from Duality: Gravity. 4 (6. In what follows we restrict ourselves to lowest-weight representations. . because those have a natural interpretation in the limit of large anti-de Sitter radius in terms of Poincar´ representations. but those will be similar and need not to be discussed separately. Unless this bound supersedes (6. Each point corresponds to the spherical harmonics of S D−1 : traceless.

1 state. s± .B. To be precise. and a “mixed” representation. For given eigenvalue E. For the bosons we consider the spherical harmonics.a1 ···al . which 2 will simply move to j = 1 . One is the symmetric traceless tensor representation (corresponding to the spherical harmonics on S D−2 ) denoted by l-rank tensors Y A1 ···Al . de Wit: Supergravity 101 Table 17. Two generic SO(D − 1) representations. representation D D=4 D=5 Y A1 ···Al l(l + D − 3) s=l s± = 1 l 2 Y B. Armed with this information it is straightforward to ﬁnd the decompositions of the spinor representation of the anti-de Sitter algebra. spanned by l-rank traceless. These states are all shown in Figure 1. as we explained in the text. for general D and for the speciﬁc cases of D = 4.a1 ···al . In the following we consider a number of representations of SO(D − 1) that exist for any dimension. the states decompose into the state of the highest spin generated by the traceless symmetric product of E − E0 + operators Ma and a number of lower-spin descendants. 2 That implies that every point with spin j in Figure 1 generates two points with spin j ± 1 . 2 . One simply takes the direct product of the spinless representation with a spin. symmetric tensors Y a1 ···al . 5. spanned by mixed tensors Y b. The result of this is shown in Figure 2.A1 ···Al of rank l + 1 (which is not independent for D = 4). For D = 4 these representations are characterized by an integer spin s. with the exception of points associated with j = 0. J . For D = 5 there are two such numbers. for general D. In a similar Table 18 we list the value of J 2 for the irreducible symmetric tensor-spinors. Multiplying such tensors with the vector representation gives rise to two of these representations with rank l ± 1. denoted by Y α. We list the corresponding eigenvalues of the quadratic Casimir operator J 2 . They are symmetric l-rank tensor spinors that vanish upon contraction by a gamma matrix and appear when taking products of spherical harmonics with a simple spinor. and the representation spanned by mixed tensors Y B. Table 17 lists the value of J 2 for these representations. The simplest case is the one where the vacuum has no spin (J = 0).A1 ···Al (l + D − 4)(l + 1) s=l s± = s∓ + 1 = 1 (l + 1) 2 The full lowest-weight representation can now be constructed by acting with the raising operators on the ground state |E0 . all states of energy E = E0 +n are constructed by an n-fold product of creation + operators Ma In this way one obtains states of higher eigenvalues E with higher spin.

The Casimir eigenvalue for this representation equals C2 (spinless singleton) = − 1 (D + 1)(D − 3) .30) The excited state then constitutes the ground state for a separate irreducible spinless representation.29) For E0 = 1 (D − 3) we have the so-called singleton representation29. but with an energy equal to E0 +2 or E0 +1.28) so that unitary of the representation requires the inequality. the spinless and the spinor representations that we have constructed so far are not necessarily irreducible. representation J2 D=4 D=5 Y α. The eigenvalues of the quadratic SO(D−1) Casimir operator J 2 for the symmetric tensor-spinor representation spanned by tensors Y α. where 2 we have only one state for each given spherical harmonic. In the context of the oscillator method. 29 The singleton representation was ﬁrst found by Dirac [86] in 4-dimensional anti-de Sitter space and was known as a “remarkable representation”. which. but now with E0 = 1 (D+1). (6. (6. To see this consider the excited state that has the same spin content as the ground state.27) This leads to 2E0 + 3 − D = 1 2 − Ma |E0 + 2.a1 ···al for general dimension D and for the speciﬁc cases D = 4. and compare their value for the Casimir operator with that of the corresponding ground state. 2 (6. spinless 2 . singletons in anti-de Sitter spaces of dimension D = 4 are called “doubletons” [87]. . In this way we ﬁnd for the scalar − E0 (E0 − D + 1) = (E0 + 2)(E0 − D + 3) + Ma |E0 + 2. Gauge Theory and Strings Table 18. respectively. 2 has the same value for C2 . spinless 2 . 5.102 Unity from Duality: Gravity. 4 (6. not surprisingly. which we will refer to later. In these lectures we will only use the name singleton to denote these remarkable representations.a1 ···al l(l + D − 2) + 1 (D − 1)(D − 2) 8 s=l+ 1 2 s± = s∓ − 1 = 1 l 2 2 However. for the scalar and spinor representations. E0 ≥ 1 (D − 3) .

The existence of the singletons was ﬁrst noted by Dirac [86].33) For E0 = 1 (D − 2) we have the spinor singleton representation.. For the spinor representation the value of the Casimir operator equals 1 C2 (spinor singleton) = − 8 (D + 1)(D − 2) . as is obvious in Figure 3. In contradistinction. associated with two spherical harmonics. which simply move to j = 1.34) Note that in D = 4. de Wit: Supergravity For the spinor representation one ﬁnds a similar result. To see this. the singleton spectrum is diﬀerent as the states have a single energy eigenvalue for any given value of the spin. note that Poincar´ representations correspond e e to plane waves which are decomposable into an inﬁnite number of spherical harmonics. wave functions that constitute singleton representations do not depend on the radius of the anti-de Sitter spacetime and can be regarded as living on the boundary. Now the situation is more complicated. (6. These representations are characterized by the fact that they do not exist in the Poincar´ limit. irrespective of the size of the spatial momentum (related to the energy eigenvalue). as the resulting multiplet contains states of spin lower than that of the ground state. Figure 1. each point in Figure 1 now generates three points. for given spin. (6. which again 2 consists of just one state for every value of the total spin. one is dealing with an inﬁnite. In principle. (6. To obtain the spin-1 representation one can take the direct product of the spinless multiplet with a spin-1 state. 2 (6. however. The result of taking the product is depicted in Figure 4. which is just what one obtains in the limit of vanishing energy increments for the generic spectrum shown in. e. An exception are the spinless points. spinor 2 103 . associated with rank-j ±1 tensors as well as mixed tensors of rank j + 1 (so that l = j). That means that.32) so that one obtains the unitarity bound E0 ≥ 1 (D − 2) . both singleton representations have the same eigenvalue of the Casimir operator.31) As the value for J 2 are the same for the ground state and the excited state one readily derives − 2E0 − D + 3 = Ma |E0 + 1. E0 (E0 − D + 1) + J 2 = (E0 + 1)(E0 − D + 2) + J 2 − + Ma |E0 + 1. Consequently. continuous tower of modes.B.g. This procedure can be . spinor 2 .

l − 1 2 E0 (E0 − D + 1) + l(l + D − 3) . States of the spinor representation in terms of the energy eigenvalues E and the angular momentum. In that case we ﬁnd C2 = = so that E0 − l − D + 3 = 1 2 − Ma |E0 + 1. 2.j 0 1 2 3 4 5 6 b b b s s s s s b b s s s s b b s s s b s s b s Fig. The small circles denote the original spinless multiplet from which the spinor multiplet has been constructed by a direct product with a spinor. assuming that the ground state has l ≥ 1. Hence we consider the states with E = E0 + 1 and j = l − 1.104 Unity from Duality: Gravity. We compare the value of the Casimir operator for the ﬁrst excited states with minimal spin to the value for the ground state speciﬁed in (6.25). we investigate whether this representation can become reducible for special values of the ground state energy. l − 1 2 − (E0 + 1)(E0 − D + 2) + (l − 1)(l + D − 4) − Ma |E0 + 1. extended directly to ground states that transform as a spherical harmonic Y a1 ···al . Along the same lines as before.36) . (6. (6. the half-integer values for j = l + 1 denote that we 2 are dealing with a symmetric tensor-spinor of rank l.35) . Gauge Theory and Strings E 6 E0 + 5 E0 + 4 E0 + 3 E0 + 2 E0 + 1 E0 b b b s s s s s s .

the state |E0 + 1. however. 3. This can be interpreted as the result of a gauge symmetry. as is explained in the text. Because these representations have a smooth Poincar´ limit they are not singletons and can therefore be regarded e as massless representations. which denotes the spin-0 ground state energy.j 0 1 2 3 4 5 6 s c s c s c s c s c 105 Fig.1 singleton. Hence massless representations with spin l ≥ 1 are characterized by E0 = l + D − 3 .1 singleton ground state has an energy which is one 2 2 half unit higher.38) .1 singleton representations.37) When E0 = l + D − 3. The spin. l − 1 is itself the ground state of an irreducible multiplet. which decouples from the original multiplet together with its corresponding excited states. is equal to E0 = 1 (D − 3). Therefore we establish the unitarity bound E0 ≥ l + D − 3 (l ≥ 1). 2 It is obvious that singletons contain much less degrees of freedom than a generic local ﬁeld. the circles the states of the spin. The spin-0 and spin.B. The value of E0 . (l ≥ 1) (6. The solid dots indicate 2 the states of the spin-0 singleton. de Wit: Supergravity E 6 E0 + 5 E0 + 4 E0 + 3 E0 + 2 E0 + 1 E0 s . (6.

The double-occupancy points exhibit the structure of a spin-0 multiplet with ground state energy E0 + 1. We return to this issue at the end of this section. at least for general dimension. States of the spin-1 representation in terms of the energy eigenvalues E and the angular momentum j. but this requires further knowledge of the various representations of the rotation group. 4. The above arguments can be easily extended to other grounds states. However. the identiﬁcation of massless representation is somewhat ambiguous. as we already discussed. Gauge Theory and Strings E 6 E0 + 5 E0 + 4 E0 + 3 E0 + 2 E0 + 1 E0 s s . 5 dimensions this information is readily available. This multiplet becomes reducible and can be dropped when E0 = D − 2. as is explained in the text. in D = 4. The remaining points then constitute a massless spin-1 multiplet. shown in Figure 5. For these particular values the quadratic Casimir operator acquires a minimal value equal to C2 (massless) = 2(l − 1)(l + D − 3) .j 0 1 2 3 4 5 6 s f s s s f s s f s s f f s s f s s s Fig. indicated by the circle superimposed on the dots and states transforming as mixed tensors (with l = j) denoted by a . For a spin-s ground state in 4 spacetime dimensions we . (l ≥ 1) (6. Observe that there are now points with double occupancy. This is outside the scope of these lectures. For certain other cases.106 Unity from Duality: Gravity.39) We recall that this result is only derived for l ≥ 1.

j 0 1 2 3 4 5 6 s s s s s s s s s s s 107 Fig. (6. States of the massless s = 1 representation in terms of the energy eigenvalues E and the angular momentum j. because here the rotation group factorizes into two SU (2) groups. 1 conformal ﬁelds.42) . The equation corresponding to (6. When the bound is saturated we obtain a massless representation. It turns out that this result applies to all spin-s representations.37). This is a 2 special property for D = 4 dimensions.41) C2 = 2(s2 − 1). 5. even to s = 0. Now E0 is no longer arbitrary but it is ﬁxed to E0 = D − 2. 2 E0 ≥ s + 1 . s− ) with s± ≥ 1 . The case of D = 5 requires extra attention. de Wit: Supergravity E 6 E0 + 5 E0 + 4 E0 + 3 E0 + 2 E0 + 1 E0 s . First let us assume the the groundstate has spin (s+ . In 2 that case we ﬁnd that the ground state energy satisﬁes the unitarity bound.40) by following the same procedure as leading to (6.39) becomes (6.B. We brieﬂy summarize some results. for which we cannot use this derivation. (s± ≥ 1 ) 2 (6. E0 ≥ s+ + s− + 2 . immediately derive the unitarity bound (for s > 1 ).

As we demonstrated . for which the E = E0 + 1 states with spin (s+ − 1 . the shortening of the representation is qualitatively similar to the shortening of BPS multiplets based on the anticommutator of the supercharges. Here we ﬁnd (6.43) For s± = 1 l these values are in agreement with earlier result. unlike in 4 dimensions. When viewed in this way. We close this section with a number of comments regarding “massless” representations and their ﬁeld-theoretic description.45) The singleton representations for s = 0. is known as the oscillator method [88]. When the bound is saturated we have again a singleton representation. A powerful method to construct the unitary irreducible representations of the anti-de Sitter algebra. Our discussion of the shortening of anti-de Sitter supermultiplets in Section 6. The corresponding values for the Casimir operator are C2 (singleton) = 3(s2 − 1) . Gauge Theory and Strings This bound is saturated for massless states. This is generically the case for arbitrary dimensions D = 4 and is thus related to the fact that the rotation group is of higher rank. The same phenomenon of multiplet shortening is well known and relevant in conformal ﬁeld theory in 1 + 1 dimensions.44) E0 ≥ 1 + s . 2 What remains to be considered are the ground states with spin (0. which is applicable in any number of spacetime dimensions and which can also be used for supersymmetric extensions of the anti-de Sitter algebra. for a recent elementary introduction to this method we refer to [80]. Here we are not striving for completeness.3 will support this point of view. just as in the earlier discussions on BPS multiplets in previous sections. 1 were already found earlier. s− − 1 ) decouples. with a large variety of spin values. (6. From the above it is clear that we are dealing with the phenomenon of multiplet shortening for speciﬁc values of the energy and spin of the representation. Mb ] commutator acquires zero or negative eigenvalues for certain values of E0 and J 2 . s). The purpose of this section was to elucidate the various principles that underlie the anti-de Sitter representations and their relation with the ﬁeld theory description.108 Unity from Duality: Gravity. There is an extensive literature on this. This phenomenon can be understood from the fact that − + the [Ma . (6. The corresponding value for 2 2 the Casimir operator is equal to C2 = (s+ + s− )2 + 2s+ (s+ + 1) + 2s− (s− + 1) − 4 . Note 2 that for D = 5 there are thus inﬁnitely many singleton representations. There are in fact alternative and often more systematic techniques for constructing the lowest-weight representations.

1 representations with the criteria 2 according to which they can be regarded as massless. where n is an arbitrary positive integer.38) does not apply so that the interpretation as a massless representation is less obvious. This anti-de Sitter representation is described by a (complex) tensor ﬁeld. it also contains the l = 0 representation with E0 = D − 3. (6. when they e do not.44) and (6.10) so that C2 = 1 (cf. We also invoke a criterion introduced by G¨naydin (see [89] and the u discussion in [90]). which are obviously massless in view of (6. 0) and energy E0 = 2 + s + n. they are called singletons.46) where m = ±g. 0) leads to a ground state with spin (s+ . it is easy to verify that the product of two spinless singletons leads to an inﬁnite series of higher spin representations that are all massless according to (6. This phenomenon takes place when some unitarity bound is saturated. Another one follows from the fact that we are dealing with a gauge ﬁeld. In that case one has representations that contain fewer degrees of freedom.13)). For the case of spin-0 or spin. However. decouple into diﬀerent irreducible representations. certain representations can. The tensor product of the singleton representations with spin (0. One of them is tied to the fact that the corresponding ﬁeld equation is conformally invariant. Therefore the concept of mass remains ambiguous.46) projects out the . Hence these representations should be regarded as massless. as we discussed at the end of Section 6. It is interesting to consider this criterion for masslessness in D = 5 dimensions. for example.1.38). de Wit: Supergravity 109 above. whose ﬁeld equation takes the form. Taking the product of two singleton representations. 0) and another one with spin (s2 . according to which every representation should be regarded as massless that appears in the tensor product of two singleton representations. s− ) and (s+ . one of these representations appears as part of the “massless” supergravity multiplet.1. This interpretation can be tested as follows. as long as E0 does not saturate the unitarity bound and a singleton representation decouples.B. for a speciﬁc value of E0 . In Table 19 we have collected a number of examples of spin-0 and spin. We have already discussed this in Section 6.42). From this equation one can show that Bµν satisﬁes (6. to which (6. For instance. On shell the equation (6. one with spin (s1 . (6. Here the example is an antisymmetric rank-(D − 2) gauge ﬁeld. which is on-shell equivalent to a scalar. the spectrum of states is qualitatively independent of the value for E0 . When these “shortened” representations have a smooth Poincar´ limit. 0) leads to ground states with spin (s. where we emphasized that the absence of mass terms in the ﬁeld equations is also not a relevant criterion for masslessness. s− ) and E0 = 2 + s+ + s− .1 repre2 sentations. In maximal gauged supergravity in 5 dimensions with gauge group SO(6). e−1 εµνρσλ Dρ Bσλ + 2im B µν = 0 . they are called massless.

91]. These anti-de Sitter supermultiplets were ﬁrst discussed in [82. 0) or (0. 1). (6. In most of the section we discuss simple supersymmetry (i. 85. Γa = 0 iσ a −iσ a 0 . .22) and obtain [H . N = 1). We substitute these deﬁnitions into (6. spin 0 0 0 0 1 2 1 2 D−3 D−1 1 1 2D − 2 5 D− 2 E0 1 2D − 1 2D 1 − 2) − 2) −2(D − 3) 0 1 − 8 D(D − 1) 1 (D2 − 15D + 32) 8 C2 1 − 4 D(D 1 − 4 D(D type conformal scalar conformal scalar ∈ singleton × singleton (D − 2)-rank gauge ﬁeld conformal spinor ∈ singleton × singleton degrees of freedom corresponding to spin (1. .44)).110 Unity from Duality: Gravity. 84. depending on the sign of m. but at the end we turn to more general N . . 2.e. For deﬁniteness we discuss the case of 4 spacetime dimensions with a Majorana supercharge Q.22). 6. (6. Some unitary anti-de Sitter representations of spin 0 and 1 which are 2 massless according to various criteria.47) and write the Majorana spinor Q in the form Q= qα εαβ q β . 3 . β. We start from the (anti-)commutation relations already presented in (6. and the corresponding values for E0 and C2 . We choose conventions where the 4 × 4 gamma matrices are given by Γ0 = −i1 0 0 i1 . the indices α. qα ] = 1 − 2 qα . This allows us to make contact with the material discussed in Section 6. From this one derives that E0 = 3 (a second solution with E0 = 1 violates the unitarity bound (6.1. Gauge Theory and Strings Table 19.3 The superalgebras OSp(N |4) In this section we return to the anti-de Sitter superalgebras.. are two-component spinor indices and a the σ are the Pauli spin matrices. a = 1. .48) † where q α ≡ qα .18) and (6.

with energy E0 + 1 . We see that the operators qα and q α are lowering and raising operators. both with energy E0 + 1 . characterized by the fact that it is annihilated by the lowering operators qα .50) In principle we can now choose a ground state and build the whole representation upon it by applying products of raising operators q α . As we will 2 see below. s . We assume that there exists a lowestweight state |E0 .2 pair of ground states q α |E0 . respectively.1 singleton. However. q β } = 1 α 2q 111 . respectively.1. s which have both spin s and which have energies E0 and E0 + 1. q α ] = {qα . 0 with ground-state energies E0 and E0 + 1. we study unitary irreducible representations of the OSp(1|4) superalgebra. we only have to study the antisymmetrized products of the q α . 2 For s ≥ 1 we are in the generic situation. The s = 0 case is special since it contains less anti-de Sitter representations than the generic case. consisting of a spin-0 and a spin. Observe that the relative sign between H and J · σ in the third (anti)commutator is not arbitrary but ﬁxed by the closure of the algebra. s = 0 . so that E0 > 1 . s and q [α q β] |E0 . − Ma (σ a σ 2 )αβ . For E0 = 1 the 2 2 multiplet degenerates and decomposes into a super-singleton. these states can be described by the scalar ﬁeld A. that we studied in Section 6. q α |E0 . s .49) where we have deﬁned the (hermitean) angular momentum operators Ja = 1 − 2 i εabc M bc . respectively. We obtain the ground states 2 |E0 . 0 .B. s . Obviously. q α |E0 . qβ } = {q α . the bounds for E0 that we derived in the previous sections should be respected. They change the energy of a state by half a unit. These ground states are |E0 . qα |E0 . q β } = {qα . + Products of the Ma simply lead to the higher-energy states in the antide Sitter representations of given spin that we considered in Section 6. s . By restricting ourselves to the antisymmetrized products of the q α we thus restrict ourselves to the ground states upon which the full anti-de Sitter representations are build. There 1 is one spin. and another spinless supermultiplet with 2 E0 = 3 .49). 0 and q [α q β] |E0 . There are two more (degenerate) ground states. (H 1 + J · σ)α β . de Wit: Supergravity [H . It includes the spinless states |E0 . Let us brieﬂy discuss these representations for diﬀerent s. which decompose into the ground states 2 with spin j = s − 1 and j = s + 1 . the pseudoscalar ﬁeld B and the spinor ﬁeld ψ of the scalar chiral supermultiplet. (6. because the + symmetric ones just yield products of the operators Ma by virtue of (6. + Ma (σ 2 σ a )αβ . (6. s and q [α q β] |E0 .2. In analogy to the bosonic case. 2 2 .

as follows by straightforward calculation. this matrix must be positive deﬁnite.52) (J · σ)2 + (J · σ) = s(s + 1)1 . We now discuss this in a general way analogous to the way in which one discusses BPS multiplets in ﬂat space. This is the bound (6. s|{qα . but in order to be positive deﬁnite the eigenvalue E0 of H must be big enough to compensate for possible negative eigenvalues of J · σ. In that case we must be dealing with a massless multiplet. as one can verify by inserting a complete set of intermediate states between the operators qα and q β in the matrix element on the left-hand side. so that it is (4s + 2)-by-(4s + 2). s E0 . the expression on the right-hand side of (6.2.51) to be positive deﬁnite.40). The ground-state energy for anti-de Sitter multiplets corresponding to the scalar ﬁeld A. i. we consider the matrix elements of the operator qα q β between the (2s + 1)-degenerate ground states |E0 . we note that J · σ satisﬁes the following identity (6.112 Unity from Duality: Gravity. There is an extended literature on this. Namely. The corresponding multiplets are then again called massless. s · (6.51) has zero eigenvalues so that there are zero-norm states in the multiplet which decouple. s| qα q β |E0 . the right-hand side is manifestly hermitean as well. s E0 . Hence in order for (6. for instance [87. The ground state with s = 1 and E0 = 3 leads to the massless 2 2 vector supermultiplet in 4 spacetime dimensions. q β }|E0 .51) This expression constitutes an hermitean matrix in both the quantum numbers of the degenerate groundstate and in the indices α and β. This shows that J · σ has only two (degenerate) eigenvalues (assuming s = 0. if E0 = s + 1.e. To determine its eigenvalues. namely s and −(s + 1). Because we assume that the representation is unitary.53) If the bound is saturated. s = = E0 .2 . the pseudo-scalar ﬁeld B and . whose applicability is extended to 1 spin. As we already mentioned one can also use the oscillator method to construct the irreducible representations. there can be situations in which states decouple so that we are dealing with multiplet shortening associated with gauge invariance in the corresponding ﬁeld theory. s|(E0 1 + J · σ)α β |E0 . for s ≥ 1 · 2 (6. Gauge Theory and Strings As in the purely bosonic case of Section 6. E0 must satisfy the inequality E0 ≥ s + 1 . Obviously. 92]. so that the above equation is not trivially satisﬁed).1 for the chiral supermultiplet. where the latter is again regarded as a (4s + 2)-by-(4s + 2) matrix. The reader may consult. Armed with these results we return to the masslike terms of Section 6.

in view of the fact that it contains a tensor gauge ﬁeld. Comparison with the ﬁeld equations (6. i. For toroidal compactiﬁcations the massive Kaluza-Klein states belong to BPS multiplets whose central charges are the momenta associated with the compactiﬁed dimensions. 2 The Casimir operator therefore takes the values C2 (A) C2 (B) C2 (ψ) = E0 (E0 − 3) .6) shows (for g = 1) that we obtain the correct contributions provided we make the identiﬁcation E0 = m + 1. The reason is that the underlying supergravity multiplet in higher dimensions is shortened because it is massless. C2 (A) + 2 C2 (B) + 2 C2 (ψ) + 3 2 = (E0 − 1)(E0 − 2) . are equal to E0 . The spinless representation with 1 E0 = 2 can be described by a scalar ﬁeld. and the second spinless representation with E0 = 3 by a rank-2 tensor ﬁeld. but also the inﬁnite tower of massive Kaluza-Klein states should comprise shortened supermultiplets. an alternative ﬁeld representation for describing this supermultiplet. = (E0 + 1 )(E0 − 5 ) + 3 · 2 2 4 (6. = (E0 − 1)2 . = E0 (E0 − 1) . de Wit: Supergravity 113 the Majorana spinor ﬁeld ψ.54) For massless anti-de Sitter multiplets.2 representation with 5 E0 = 2 by a spinor ﬁeld. (6. and this could account for the unusual ground state energy for the spinor representation. When compactifying to an anti-de Sitter ground state with supersymmetry the massless supermultiplets remain shortened by the same mechanism. the spin. The Lagrangian for the tensor supermultiplet is not conformally invariant in 4 dimensions.41). For nontrivial compactiﬁcations that correspond to supersymmetric . = (E0 + 1)(E0 − 2) .e. From Kaluza-Klein compactiﬁcations of supergravity one can deduce that there should also exist shortened massive supermultiplets. so we present the value for C2 − 2(s2 − 1) for the three multiplets. we know that the quadratic Casimir operator is given by (6. E0 +1 and E0 + 1 . the above result remains the same under the interchange of A and B combined with a change of sign in m (the latter is accompanied by a chiral redeﬁnition of ψ). When E0 = 2 there exists.B. in principle. Observe that we could have made a slightly diﬀerent identiﬁcation here.55) The terms on the right-hand side are not present for massless ﬁelds and we should therefore identify them somehow with the common mass parameter m of the supermultiplet. it should be regarded as massless. We have not constructed this supermultiplet in anti-de Sitter space. respectively.

(6. + Ma (σ 2 σ a )αβ δ ij . Hence the generic N -extended representations decompose into ordinary anti-de Sitter representation whose ground states have energy E0 + 1 n and which 2 can be written as (6. the massive Kaluza-Klein states must be shortened according to the mechanism exhibited in this section. which reads j β j β {qαi .51)). . N . taken in the space of ground state conﬁgurations (cf. qβj } {q αi . In a similar way one can determine the eigenvalues for T · Σ by noting that it satisﬁes a polynomial . . As it turns out.56). 4).49) is in the third one.59) q [α1 i1 · · · q αn in ] |E0 . t · Here the antisymmetrization applies to the combined (αi) labels. To exhibit this we generalize the previous analysis to the N -extended superalgebra. q βj } = = − Ma (σ a σ 2 )αβ δij . which are mutually anticommuting. Excited states are generated by application of the q αi . (6.57) The construction of lowest-weight representations proceeds in the same way as before.56) where T are the hermitean 1 N (N − 1) generators of SO(N ) which act 2 on the supercharges in the fundamental representation. (6. One starts with a ground state of energy E0 which has a certain spin and transforms according to a representation of SO(N ) which is annihilated by the qαi . with q αi = (qαi )† with i = 1. The supercharges now carry an extra SO(N ) index and are denoted by qαi and q αi . . the rotation generators and the R-symmetry generators.58) qαi |E0 . denoted by OSp(N. t = 0 .114 Unity from Duality: Gravity. q βj } = δi δα H + δi J · σα β + δα T · Σi j . Gauge Theory and Strings anti-de Sitter ground states. The last two anticommutators are given by {qαi . As before the unitarity limits follow from the separate limits on the anti-de Sitter representations and from the right-hand side of the anticommutator (6. namely the Hamiltonian. generated by the hermitean matrices Σ. Denoting the SO(N ) representation by t (which can be expressed in terms of the eigenvalues of the Casimir operators or Dynkin labels). we have (6. the analysis is rather similar. s. . with exception of the combination that leads to the opera+ tors Ma which will generate the full anti-de Sitter representations. We have already determined the possible eigenvalues of J ·σ which are equal to s or −(s+1). s. The singleton multiplets decouple from the Kaluza-Klein spectrum. which decomposes into three terms. Therefore it follows that there must exist shortened massive representations of the extended supersymmetric anti-de Sitter algebra. The most relevant change to the (anti)commutators (6.

E0 + s + 1. t = 0 . certain anti-de Sitter representations must decouple. For instance.52) with coeﬃcients determined by the Casimir operators. [87. Incorporating also the possibility that s or t vanishes. which correspond to N = 4. either by changing the number of supersymmetries or by changing the spacetime dimension.56) in the space of degenerate ground state conﬁgurations has the following six eigenvalues: E0 + s − t. the combined result takes the following form. E0 − s − t − 1. Group-theoretically they are of interest because their products lead to the massless and massive representations that one encountered in the Kaluza-Klein context.B. t ≥ 1 · 2 (6. Combining these results we ﬁnd that the right-hand side of (6. 92]). 1 or t + 1. The singletons decouple from the KaluzaKlein spectrum. Before closing the section we want to return to the remarkable singleton representations. namely the singleton representations in D = 5 and 7 anti-de Sitter space. In [94] the reader may also ﬁnd a . In the Poincar´ limit these representations become e all massless. 2 for s = 0 . unless t = 0. Here the issue is whether the singletons play only a group-theoretic role or whether they have also a more dynamical signiﬁcance. Another theme addresses the connection between singletons and higher-spin theories. The ground states with s = 0 and E0 = t deﬁne massive shortened representations of the type that appear in Kaluza-Klein compactiﬁcations [84]. We refrain from speculating about these questions and just refer to some recent papers [93–95]. This equation shows that the eigenvalues of T · Σ take the values −t. so that in the generic case where s and t are nonvanishing. E0 − s or E0 + t + 1. E0 ≥ 1 + s+ t. E0 + s + t + 1. E0 ≥ 1 + s + t E0 ≥ 1 + s E0 ≥ t for s ≥ 1 .61) Whenever one of these bounds is saturated. e. de Wit: Supergravity 115 matrix equation such as (6. There is an extended literature to which we refer the reader for applications and further details. precisely because they are related to boundary degrees of freedom.g. D = 4 supersymmetric gauge theories and the chiral (2. for N = 3 we derive. 0) tensor multiplet in D = 6 dimensions. Two prominent examples were noted (see. −(T · Σ)3 + 2(T · Σ)2 + (t2 + t − 1)(T · Σ) = t(t + 1) 1 .60) where T 2 = t(t + 1) 1. Obviously these techniques can be extended to other cases. (6. All these eigenvalues must be positive. we derive the unitarity bound. t ≥ 1 . 2 2 for s ≥ 1 . Long before the formulation of the AdS/CFT correspondence it was realized that supersingleton representations could be described by conformal supersymmetric ﬁeld theories on a boundary.

2) Solutions to this equation are called conformal Killing vectors. A space that has the maximal number of isometries is called maximally symmetric. Note that the above equation is the traceless part of (7. where ξ = 0 one then proves that (D − 2)Dµ Dν ξ and Dµ Dµ ξ are determined in terms of lower derivatives. For D = 2 there can be inﬁnitely many conformal Killing vectors. the Killing vector is thus fully determined by its value at that point and the values of its ﬁrst derivatives (which are antisymmetric in view of (7.1). The conformal Killing vectors that are not isometries are thus characterized by a nonvanishing ξ = Dµ ξ µ . D (7. Dµ ξν + Dν ξµ = 2 gµν Dρ ξ ρ . For general dimension D > 2 there are at most 1 (D + 1)(D + 2) conformal 2 Killing vectors. This suﬃces to derive the maximal number of conformal Killing vectors quoted above for D > 2. Continuous isometries are generated by so-called Killing vectors.1) is.116 Unity from Duality: Gravity. D (7. The corresponding superalgebra is then the anti-de Sitter superalgebra.1) The maximal number of linearly independent Killing vectors is equal to 1 2 D(D + 1). 7 Superconformal symmetry Invariances of the metric are known as isometries. In the next section we will move to a discussion of superconformal symmetries. A weaker condition than (7. We draw the attention of the reader to the fact that in Section 7. Both ordinary and conformal Killing vectors generate a group. First one shows that Dµ Dν ξρ = Rνρµ σ ξσ − 1 gµν Dρ ξ − gρµ Dν ξ − gρν Dµ ξ . Altogether there are thus 1 D(D + 1) initial conditions to be ﬁxed and 2 they parametrize the number of independent Killing vectors. When expanding about a certain point on the manifold. These result can be derived as follows.1)). (7. which are based on the same anti-de Sitter algebra. D will always denote the spacetime dimension of the superconformal theory. For conformal Killing vectors.3) For Killing vectors (which satisfy ξ = 0) this result implies that the second derivatives of Killing vectors are determined by the vector and its ﬁrst derivatives. . but in spacetime dimension D + 1. Gauge Theory and Strings summary of some useful results about singletons as well as an extensive list of references. satisfying Dµ ξν + Dν ξµ = 0 .

µ spacetime translations (P ) ξP µ xν Lorentz transformations (M ) ν ξµ = (7. assuming the line element.5) However.4) ΛD xµ scale transformations (D) (2 xµ xν − x2 η µν )ΛKν conformal boosts (K). δx = f+ (x + t) + f− (x − t) . R)×SL(2. where {µρν } remains the same for both spaces and all other components vanish. for D = 2 there are inﬁnitely many conformal Killing vectors. ds2 = gµν dxµ dxν + dy 2 . K K 2 (7.6) The fact that. 2) ∼ (SL(2. {yyy } = −y −1 . The corresponding diﬀeomorphisms can be characterized in terms of two independent functions f± and take the form. = This follows directly by writing out the inﬁnitesimal transformations (7. Flat Minkowski spacetime has the maximal number of conformal Killing vectors.4) for the linear combinations x ± t. The corresponding expressions for the curvature components are Rµνρ σ = D σ Rµνρ σ + 2y −2 gρ[µ δν] . for D ≥ 3 the anti-de Sitter and the conformal group coincide for dimensions D + 1 and D. δt = f+ (x + t) − f− (x − t) .8) {µyν } = y −1 gµν . y2 (7. The case of D = 2 is special because in that case the above transformations generate a semisimple group. The above conformal Killing vectors generate the group K SO(D. respectively. which decompose as follows. x t δ(x ± t) = (ξP ± ξP ) + (ΛD ∓ xt )(x ± t) + 1 (Λx ∓ Λt )(x ± t)2 . which is responsible for the lack of invariance of the line element of the original D-dimensional space. Obviously K ξ = D(ΛD +xµ Λµ ). µν = − νµ . 2). µ Here ξP . This is the same group as the anti-de Sitter group in D+1 dimensions. a restriction that is mainly relevant when considering supersymmetry.2).7) so that the right-hand side of (7. R))/Z2 . SO(2. {µνy } = −y −1 δµ . (7. de Wit: Supergravity 117 In what follows we choose a Minkowski signature for the D-dimensional space. ΛD and Λµ are constant parameters. corresponding to two copies of the Virasoro algebra. It is straightforward to derive the nonvanishing Christoﬀel symbols for this extended space ν (7. .B. can be cancelled by a scale transformation of extra coordinate y. can be clariﬁed by extending the D-dimensional spacetime parametrized by coordinates xµ with an extra (noncompact) coordinate y.

For D = 2 these vectors generate a ﬁnite subgroup of the inﬁnite-dimensional conformal group. This setting is relevant for the adS/CFT correspondence and there exists an extensive literature on this (see. σ y −2 δν . the conformal Killing vectors generate asymptotic symmetries.11) Near the boundary where y is small.e. y) = ξ µ (x) − y2 µ ∂ ξ(x) . This identiﬁcation of the generators is discussed in more detail in the next section. Straightforward calculation yields. e.10) The condition Dµ Dν ξ = 0 holds for the conformal Killing vectors (7.g. which associated with the SO(1. Nevertheless. 2D+1 = y 2 2D + (y ∂y )2 − D y ∂y .25)). the ﬁelds can be approximated by y ∆ φ(x). (7.14)). g = 1 in (3. In terms of the ground state energy E0 of the anti-de Sitter representation. ξ µ (x. 2D ξ y (x. [28. y) = y ξ(x) . This was the reason why we adopted a positive signature in the line element (7. Gauge Theory and Strings Rµyρ y Ryνy σ = = y −2 gµρ . near the boundary [77] of the space (y ≈ 0). Subsequently one can show that the D-dimensional conformal Killing vectors satisfying Dµ Dν ξ = 0 can be extended to Killing vectors of the (D + 1)-dimensional space. we have C2 = E0 (E0 − D) (observe that we must replace D by D + 1 in (6. Such a phenomenon was ﬁrst analyzed in [96]. and only this group can be extended to isometries of the (D+1)-dimensional space. D (7. Also the relation between the D’Alembertians of the extended and of the original D-dimensional spacetime is relevant in this context. . which shows that we have the identiﬁcation ∆ = E0 or ∆ = D − E0 .4). This identiﬁcation is somewhat remarkable in view of the fact that E0 is the energy eigenvalue associated with the SO(2) generator of the anti-de Sitter algebra and not with the noncompact scale transformation of y. We may compare this to solutions of the Klein-Gordon equation in the anti-de Sitter space. 1) eigenvalue.7) for the coordinate y. for which we know that the D’Alembertian equals the quadratic Casimir operator C2 . 97–101]. and also the lectures presented at this school).9) With these results one easily veriﬁes that the curvature tensor of the (D+1)dimensional extension of a ﬂat D-dimensional Minkowski space is that of an anti-de Sitter spacetime with unit anti-de Sitter radius (i.118 Unity from Duality: Gravity. (7.

D − 1. The decomposition of the conjugate spinor is somewhat subtle.5 (see. The algebra associated with SO(D. [Ka . . For the bosonic generators which generate the group SO(D. Sα ¯ ¯ ¯ ¯ (7. Pc ] [Mab .22). Γa = Γa 0 0 −Γa ΓD = 0 1 1 0 Γ− = 0 1 −1 0 . so that A = −.13) To obtain the superextension (for D ≤ 6) one must ﬁrst extend the spinor representation associated with the D-dimensional spacetime to incorporate two extra gamma matrices ΓD and Γ− . D. 2(Mab + ηab D) . . 0. and D generators Ka of the conformal boosts.B. Qα and we deﬁne an extended set of gamma matrices ΓA by. D generators Pa of the translations. [D. 2) we have MD− Mab MDa M−a −→ D .1 The superconformal algebra 119 From the relation between the conformal and the anti-de Sitter algebra one can determine the superextension of the conformal algebra generated by the above conformal Killing vectors. (7. 2) was given in (6. −2 ηc[a Kb] . Kc ] [Pa . (7.4) and corresponds to the following commutation relations [D. 1 D(D−1) generators 2 Mab of the Lorentz transformations. −→ −→ 1 2 (Pa 1 2 (Pa − Ka ) . Pa ] [Mab . According to the discussion in Section 2. which we extend with two extra index values. Pb ] = = = = Ka . Mcd ] [D. Table 9) this requires a doubling of the spinor charges. −→ Mab . . . . Sα ) . de Wit: Supergravity 7. Ka ] [Mab . Q → Q = (Qα . to make contact with the Majorana condition employed for the anti-de Sitter algebra. . in particular.15) The new charges Sα generate so-called special supersymmetry transformations [27]. 0. Pb ] [Ka . −2 ηc[a Pb] . (7. + Ka ).4) and (6. 1. We start from a D-dimensional spacetime of coordinates carrying indices a = 0. Kb ] = 0 . 4 η[a[c Md]b] . . . D − 1.5)) we make a diﬀerent decomposition than the one that led to (6. (6. Mab ] = = = = −Pa .14) Q→Q= . .18) and (6. In comparison to the anti-de Sitter algebra and superalgebra (cf.12) Here we distinguish the generator D of the dilatations. 1.

Qα ] = 1 (QΓab )α . the other (anti)commutation relations listed above remain unchanged. ¯ ¯ [Mab . Those multiplets are in one-to-one correspondence with the anti-de Sitter supermultiplets. Note also that the numbers of bosonic and fermion generators do not match.18) Here we are assuming the same gamma matrix conventions as in the beginning of Section 3. (7. leading to shortened supermultiplets. 2 ¯α ] = 1 Sα . Sβ } = − 1 Γab Mab + ηαβ D.120 Unity from Duality: Gravity. 2 αβ (7. αβ ¯β } = −Γa Ka . The validity of the remaining Jacobi identities. Qα ] = −(SΓa )α . Qα ] = 1 (QΓab )α . S 2 ¯ ¯ [Pa .16) ¯ ¯ {Q. this mismatch will in general remain when including the R-symmetry generators. Qβ } = −Γa Pa . The results given so far suﬃce to discuss the most salient features of the superconformal algebra and henceforth we will be ignoring the contributions of the R-symmetry generators. Gauge Theory and Strings The anticommutation relation for the spinor charges follows from (6. 2 ¯ α ] = − 1 Qa .16) may have zero eigenvalues. From the results quoted in the previous section.5) and can be written as ¯ {Q. S} 1 − 2 Γab Mab − D −Γa Ka − 1 Γab Mab 2 +D −Γ Pa a The nonvanishing commutators of the spinor charges with the bosonic generators read ¯ ¯ [Mab . S} . ¯ [D. the bosonic subalgebra will be the sum of the conformal algebra and the R-symmetry algebra. As before. The R-symmetry can be identiﬁed from Table 9 and the corresponding generators will appear on the righ-hand side of the {Q. Q} = = or ¯ {Qα . {Sα . commutators with the R-symmetry generators must be speciﬁed. which are cubic in the fermionic generators. In addition. ¯ [D. (7. up to D = 6. Its . Q 2 ¯ ¯ [Ka . we know that. Q} {Q. Q} ¯ {S. but those follow from the Rsymmetry assignments of the supercharges. S} anticommutator. requires in general the presence of the R-symmetry charges.17) ¯ {S. The above (anti)commutators satisfy the Jacobi identities that are at most quadratic in the fermionic generators. the matrix on the right-hand side of (7. Sα ] = −(QΓa )α . S αβ ¯ {Qα .

20) ∂xa so that φ(x) can be written as exp(xa Pa )φ(0). In principle. de Wit: Supergravity 121 eigenvalues are subject to certain positivity requirements in order that the algebra is realized in a positive-deﬁnite Hilbert space. Observe that there exists a subgroup of the conformal group that leaves a point in spacetime invariant and choose. (7.B. it follows that S(g. and ˆ ˆ ˆ O(g.22) [a 2ΛK xb] + ˆ Mab + ΛD − 2Λa xa K ˆ D+ ˆ Λa K a K φ(x) .4) it then follows that the corresponding stability group of this point is generated by the generators M of the Lorentz group. we are describing ﬂat space as a coset manifold.20). x) = exp(−y a Pa ) S(g. From (7. Combining (7. D and ˆ a . Generic ﬁelds are thus assigned to representations of this subgroup. by a suitable translation. where ξ a denotes the conformal Killing vectors parametrized in (7. (7. 0) ≈ 1 + 1 ab Mab + ΛD D + Λa K. we derive that conformal transforK 2 mations act inﬁnitesimally on φ according to δφ(x) = −ξ a ∂a φ(x) + 1 ab 2 (7. this point equal to xa = 0. we want the translation operators to act exclusively on the coordinates xa . whose generators are denoted by the matrices Mab . where the conformal group plays . The abstract algebra can be connected to the spacetime transformations (7. K On the other hand. ∂ Pa φ(x) = φ(x) .4). 0) exp(xb Pb ) . Indeed. this is an application of the theory of homogeneous spaces discussed in Section 4 and we will demonstrate this for the bosonic transformations [102].4) in ﬂat spacetime introduced at the beginning of this section.19) where S is some matrix acting on the components of φ. To see this we derive how the conformal transformations act on generic ﬁelds. the generator D of the scale transformations and the generators K of the conformal boosts. so that (the generators have been taken antihermitean). 0) must form a representation of ˆ ˆ this subgroup. Hence we conclude that the matrices S(g.19) with (7.21) Writing this out for inﬁnitesimal transformations with y a = xa + ξ a . (7. Let us assume that the action of these spacetime transformations denoted by g takes the following form on a generic multicomponent ﬁeld φ. x) φ(g −1 x) . with y = g x . a supersymmetric extension can be given in superspace. The procedure applied above is just a simple example of the construction of induced representations on a G/H coset manifold. φ(x)−→φg (x) = S(g.

2 = Dµ ab [a b] [a ¯ + ΛK eµ − ξP fµb] − 1 ¯Γab φµ + 1 ψµ Γab η . Note that the special conformal boosts do not act on the tangent space index of the vielbein.24) Up to normalization factors. There are two approaches here which lead to related results.b − ΛD eµa . 7.2 Superconformal gauge theory and supergravity In principle it is straightforward to set up a gauge theory associated with the superconformal algebra. so that the spin connection becomes a dependent ﬁeld and the Riemann tensor becomes proportional to the curvature of the spin connection ﬁeld. Explicit evaluation then shows that the invariance of the ﬂat vielbein requires the compensating tangent-space transformations. from which it follows (cf. Then one imposes a constraint on certain curvatures. This is similar to what we described in Section 3. Gauge Theory and Strings the role of the isometry group G and the stability group plays the role of the isotropy group H. the transformation rules for the gauge ﬁelds. 4 4 . δeµa ab δωµ a a = Dµ ξP − ΛD eµ + 1 ¯Γa ψµ . One is to start from a gauge theory of the conformal group.16)) that the vielbein is constant and diagonal and the connections associated with the stability group are zero. δeµa = ab eµ. In the next two sections we will discuss how one can deviate from ﬂat space in the context of the conformal group. (3. This conformal group has a priori nothing to do with spacetime transformations and the resulting theory is described in some unspeciﬁed spacetime. while the vielbein is invariant after including the compensating transformations represented by the second line of (7. (4. generators: gauge ﬁelds: parameters: P a eµ a ξP M ab ωµ ab D bµ ΛD K fµa Λa K Q ψµ S φµ η.22). Hence the metric is invariant under the conformal transformations. This approach amounts to imposing the maximal number of conventional constraints.122 Unity from Duality: Gravity.22). where we imposed a constraint on the torsion tensor (cf.23) with parameters speciﬁed by (7. (7. follow directly from the structure constants of the superconformal algebra. as established earlier. The coset representative equals exp(xa Pa ).6)). We start by associating a gauge ﬁeld to every generator. (7. which we specify below. The second approach starts from the coupling to superconformal matter and the corresponding superconformal currents.

2 K [a b] δP ( 1 ¯2 Γa 1 ) . 2 2 2 123 = Dµ η + 1 ΛD φµ − 1 fµa Γa + 1 Λa Γa ψµ . δP (ΛP )] = δS ( 1 Λ a Γa ) .B. 2 δK ( 1 η2 Γa η1 ) . 2 ¯ 2 D[µ bν] − f a e − 1 ψ[µ φν] . ν] 2 [a b] ab ac ¯ 2 ∂[µ ων] − 2 ω[µ ων] c b − 2 f[µ eν] + 1 ψ[µ Γab φν] . see. δK (ΛK )] = {δQ ( 1 ). 2 2 2 K (7. 4 [δQ ( ). δS (η)} = = = b δD ( 1 Λa ξP ηab ) + δM (ΛK ξP ) .e. 2 (7.25) Here we use derivatives that are covariantized with respect to dilatations and Lorentz transformations. K 2¯ a = Dµ − 1 ΛD ψµ − 1 eµa Γa η + 1 ξP Γa φµ . δK (ΛK )] = [δS (η). up to the commutators of two supersymmetry transformations acting on the fermionic gauge ﬁelds. de Wit: Supergravity δbµ δfµa δψµ δφµ ¯ = Dµ ΛD + 1 ΛKa eµa − 1 ξP a fµa + 1 ¯φµ − 1 ψµ η . a Dµ ξP a a = ∂µ ξP + bµ ξP − ωµ ab ξP b . which depend sensitively on the dimension and on the presence of additional generators (for D = 4. one needs Fierz reorderings to establish the closure of the algebra. K K = ∂µ ΛD . 2 4 ab = (∂µ − 1 bµ − 1 ωµ Γab )η . 2 2 4 4 a = Dµ Λa + ΛD eµ + 1 η Γa φµ . In that case. a Rµν (P ) = ab Rµν (M ) = ¯ 2 D[µ ea − 1 ψ[µ Γa ψν] . [µ ν] a 2 a ¯ 2 D[µ fν] − 1 φ[µ Γa φν] . i.26) Again we suppressed the gauge ﬁelds for the R-symmetry generators. The above transformation rules close under commutation. 2 K a δQ ( 1 ξP Γa η) . 2 4 (7. δQ ( 2 )} {δS (η1 ). As an example we list some of the commutation relations that can be obtained from (7. 2¯ 1 δM ( 1 ¯Γab η) + δD (− 4 ¯η) ..27) For completeness we also present the corresponding curvature tensors of the superconformal gauge theory. ab = (∂µ + 1 bµ − 1 ωµ Γab ) . Dµ Λa K Dµ ΛD Dµ Dµ η = ∂µ Λa − bµ Λa − ωµ ab ΛKb .25) [δP (ξP ). 2 Rµν (D) = a Rµν (K) = . δS (η2 )} {δQ ( ). for example [27]).

when imposing the torsion constraint R(P ) = 0.22). the curvature R(P ) has the interpretation of a torsion tensor. This is completely in line with the ﬁeld transformations (7. at this stage. There is a procedure to introducing a nontrivial entangling between the spacetime diﬀeomorphisms and the (internal) symmetries associated with the superconformal gauge algebra. (7. . a a a a δeµ = Dµ ξP = ∂µ ξ ν eνa − ξ ν Dν eµ + ξ ν Rµν (P ) . based on curvature constraints. there is no need for the bosonic and fermionic degrees of freedom to match. Of course. Hence. The eﬀect of this constraint is also that the P gauge transformations are eﬀectively replaced by generalcoordinate transformations. (7. where ﬂat space was viewed as a coset space. and one can impose a constraint R(P ) = 0. They also satisfy a number of Bianchi identities which are straightforward to write down. making use of the fact that there exists an inverse vielbein ea . 4 a a D[µ Rνρ] (D) + 1 R[µν (K) eρ] a − R[µν (P ) fρ] a 2 ¯ ¯ + 1 φ[ρ Rµν] (Q) − 1 ψ[ρ Rµν] (S) = 0 .124 Unity from Duality: Gravity. Here one regards the P gauge ﬁeld eµa as a nonsingular vielbein ﬁeld. as one would expect for a conventional supersymmetric theory.30) a µ where ξ µ = ξP ea . a 2 D[µ φν] − f[µ Γa ψν] .29) 4 4 At this stage. (7. the gauge ﬁelds independently transform as vectors under general coordinate transformations but these transformations have no intrinsic relation with the gauge transformations. just as in (3. In that case. This interpretation is in line with the interpretation presented in the previous section. a P -transformation takes the form of a (covariant) general coordinate transformation.6). the superconformal algebra is not related to symmetries of spacetime. so that the M gauge ﬁeld ab ωµ becomes a dependent ﬁeld. except that we are now dealing with arbitrary diﬀeomorphisms. whose inverse will be denoted by eaµ . where the P -transformations were also exclusively represented by coordinate changes. 2 ab D[µ Rνρ] (M ) + R[µν (K) eρ] + R[µν (P ) fρ] [a b] [a b] 1¯ ¯ + 4 φ[ρ Γab Rµν] (Q) + 1 ψ[ρ Γab Rµν] (S) = 0 . Gauge Theory and Strings Rµν (Q) = Rµν (S) = a 2 D[µ ψν] − e[µ Γa φν] . a ab a ¯ D[µ Rρ] (P ) + R[µν (M ) eρ] b − R[µν (D) eρ] − 1 ψ[ρ Γa Rµν] (Q) = 0 . As an example and for future reference we list the ﬁrst three identities.28) These curvature tensors transform covariantly and their transformation rules follow from the structure constants of the superconformal algebra. let us rewrite a P -transformation a µ on eµ . This is the reason why. To see this.

A maximal set of conventional constraints that achieves just that. because the constraint R(P ) = 0 is inconsistent with Q-supersymmetry. under Q-supersymmetry. At this point we are left with the vielbein ﬁeld eµa . local Lorentz transformations (M ). we have diﬀeomorphisms. local conformal boosts (K). takes the form a Rµν (P ) = ab ebb Rµν (M ) Γµ Rµν (Q) 0. Note that bµ is the only ﬁeld that transforms nontrivially on special conformal boosts and therefore acts as a compensator which induces all the K-transformations for the dependent ﬁelds. Because the constraints are consistent with all the bosonic transformations. R(D) and R(Q) can be conventional and may lead to additional dependent gauge ﬁelds fµa and φµ associated with special conformal boosts and special supersymmetry transformations.25). for reasons of covariance.25). The gauge transformations remain with the exception of the P transformations. the ﬁeld the ﬁeld ωµab will still transform under these symmetries according to (7. 0. de Wit: Supergravity 125 A constraint such as R(P ) = 0 is called a conventional constraint. as one can verify by comparing the numbers of . the precise form of the constraints is not so important.28) shows that constraints on R(M ). one should include possible modiﬁcations of the curvatures due to the changes in the transformation laws of the dependent ﬁelds. One of them is that there are potentially more conventional constraints. 0.29) and should not be constrained. Of course. local scale transformations (D). which can easily be eliminated by a ﬁeld redeﬁnition. those will not change and will still describe a closed algebra. and the bosonic conformal transformations. the gauge ﬁeld bµ associated with the scale transformations. the transformation rules of the dependent ﬁelds are determined and they may acquire extra terms beyond the original ones presented in (7. Note that Rµν (D) is a not independent as a result of the ﬁrst Bianchi identity on Rµν (P ) given in (7. because constraints that diﬀer by the addition of other covariant terms result in the addition of covariant terms to the dependent gauge ﬁelds. however. Indeed.25). because it algebraically expresses some of the gauge ﬁelds in terms of the others. We will not elaborate on the systematics of this procedure but concentrate on a number of noteworthy features. but not for Q-supersymmetry. by doing so. not close. the ﬁeld ωµab acquires an extra term beyond what was presented in (7. This is also the case for S-supersymmetry. Other than that. and the gravitino ﬁeld ψµ associated with Q-supersymmetry. which is proportional to R(Q). (7. Because R(P ) = 0 is consistent with spacetime diﬀeomorphisms. Inspection of (7. All other gauge ﬁelds have become dependent. The superalgebra will. Q-supersymmetry and S-supersymmetry.31) = = where.B.

Suppose that we couple such a rigidly superconformal matter theory in ﬁrst order to the gauge ﬁelds of conformal supergravity. eµa = δµ . see [103]). comprise the smallest massive supermultiplet whose highest spin coincides with the graviton spin.e. In this background we may have (matter) theories that are superconformally invariant under rigid transformations. Gauge Theory and Strings bosonic and fermionic degrees of freedom. Hence we write. Here ns denotes the spinor dimension. gauge invariances of the conformal 2 2 group. In order to have a consistent superconformal theory one must add additional ﬁelds (for a review. 104. A similar oﬀ-shell counting argument applies to the fermions. This theory has an obvious rigid limit. A practical way to do this makes use of the superconformal multiplet of currents [22]. eµ ≈ δµ + hµa . implies that they must constitute massive representations of the Lorentz group.126 Unity from Duality: Gravity. Hence. This leaves us with 1 D(D − 1) − 1 degrees of freedom. This is the background we considered in Section 7. The independent bosonic ﬁelds. This construction is limited to theories with Q = 16 supercharges and leads to consistent conformal supergravity theories [22. bosonic. which constitutes an irreducible representation of the Poincar´ algebra. described by (7. eµa and bµ . with the exception of the vielbein which a is equal to the ﬂat vielbein. (7. In linearized approximation the above Lagrangian is invariant under local superconformal .3 Matter ﬁelds and currents In the previous section we described how to set up a consistent gauge theory for conformal supergravity. a a i. This representation is the minimal representation that e is required for an oﬀ-shell description of gravitons in D spacetime dimensions. rank-2 tensor in D − 1 dimensions. the conformal framework is set up to reduce the ﬁeld representation to the smallest possible one that describes the leading spin without putting the ﬁelds on shell. L = Lmatter a ab µ ¯ ¯ µ +hµ θa µ + 1 ωµ Sab + bµ T µ + fµa Ua µ + ψµ J µ + φµ JS .1. Similarly. traceless. which comprise (D − 2)ns degrees of freedom after subtracting the gauge degrees of freedom associate with Q.and S-supersymmetry. which we will discuss in the next section. where all the gauge ﬁelds are equal to zero. The current θa µ is the energy-momentum tensor. 105]. which are subject to the 1 D2 − 3 D − 1 independent. The ﬁrst term denotes the matter Lagrangian in ﬂat space.32) 2 a where hµ denotes the deviation of the vielbein from its ﬂat space value.. 7. The fact that the ﬁelds can exist oﬀ the mass shell. the supermultiplet of ﬁelds on which conformal supergravity is based. comprise D2 + D degrees of freedom. corresponding 2 to the independent components of a symmetric. We close this section with a comment regarding the number of degrees of freedom described by the above gauge ﬁelds.22).

δbµ = 1 ΛKµ . (7. 1 δψµ = − 2 Γµ η. = 0.34) where we used the ﬂat vielbein to convert world into tangent space indices and vice versa. ab conformal boosts: δfµa = ∂µ Λa . we employed the notation θab = θa µ eµ b and θµ µ = θa µ eµa . Observe that the terms involving the energymomentum tensor take the form θa µ ξ a . To examine the consequences of this we need the leading (inhomogeneous) terms in the transformations of the gauge ﬁelds (cf. Obviously. but we can deﬁne a set of conserved currents by allowing an explicit dependence on the coordinates. where ξ a are the conformal Killing vectors deﬁned in (7. (7. [a b] (7. a a dilatations: δbµ = ∂µ ΛD . = 0. = 0.35) − b 2 θ[aµ xb] ∂µ (T µ + θa µ xa ) − 1 µ a 2T x − 1 µ 2 θb (xa x µ ∂µ JS + b − 1 x2 δa ) 2 ∂µ J µ µ ν 1 2 Γν J x In this result one recognizes the various components in (7. S-supersymmetry: δφµ = ∂µ η . = 0. One can verify that this leads to a number of conservation equations for the currents.4).32) under the superconformal transformations. .25)).4). ∂µ θa µ ∂µ ∂µ Ua − µ 1 µ b 2 Sab x µ Sab = 0. δωµ = ΛK δµ . de Wit: Supergravity 127 transformations. (7. δhµ = ab δµ b . = 0 . must vanish. For S-supersymmetry one can understand the expression of the current by noting that the following combination of a constant S transformation with a spacetime dependent Q-transformation with = 1 xµ Γµ η leaves the 2 gravitino ﬁeld ψµ invariant. ab a Lorentz: δωµ = ∂µ ab . for instance.33) The variations of the action corresponding to (7. = 0. = 0. K 2 Q-supersymmetry: δψµ = ∂µ . not all currents are conserved. ignoring variations that are proportional to the superconformal gauge ﬁelds and assuming that the matter ﬁelds satisfy their equations of motion.22) and in (7.B. = 0. µ ∂µ Ua µ − 1 Saµ − 1 Ta 2 2 ∂µ J µ µ ∂µ JS + 1 Γµ J µ 2 = 0. a a translations: δhµ = ∂µ ξP . ∂µ θa µ µ ∂µ Sab − 2 θ[ab] ∂µ T µ + θµ µ = 0. = 0. δhµ = −δµ ΛD .

which can be included into the currents without aﬀecting their divergence.40) the improved energy-momentum tensor is conserved. we have argued in the previous section that it is possible to choose the gauge ﬁelds associated with the generators M .36) Observe that these equations reduce the currents to irreducible representations of the Poincar´ group. the ﬁelds ωµab . It is.38) 2 2 It is conserved by virtue of the ﬁeld equations. 2 (7.39) + When χ(φ) satisﬁes D−2 ηµν ∂ 2 − ∂µ ∂ν χ(φ) . . (7. Ua µ and JS . Gauge Theory and Strings So far we have assumed that the gauge ﬁelds in (7. µ µ suppressing Sab .32) are independent. imp θµν = A 1 2 gAB (∂µ φ ∂ν φB − 1 ηµν ∂ρ φA ∂ ρ φB ) 2 (7. (7. After a partial integration. From this result it follows that locally in the target 30 A homothetic vector satisﬁes D χ + D χ A B A B = 2gAB Here we are dealing with an exact homothety. the currents θa µ . To illustrate the construction of the currents. Hence. for which DA χB = DB χA . 4(D − 1) DA ∂B χ(φ) = gAB . K and S. because we can simply repeat the analysis leading to (7. µ imp imp ∂ µ Jµ = Γµ Jµ = 0 . χA = ∂A χ is a homothetic vector30. and which can be solved by a potential χ. upon using the ﬁeld equations). We then obtain the following conditions for the improved currents. possible to introduce an improvement term. L = 1 gAB ∂µ φA ∂ µ φB . However. T µ and J µ are modiﬁed by improvement terms: terms of the form ∂ν A[νµ] . moreover it is symmetric. Ua µ and JS no longer appear explicitly but are absorbed in the remaining currents as improvement terms. to depend on the other ﬁelds. imp imp ∂ µ θµν = θ[µν] = θimp µ = 0 . This implies that. We do not have to work out their explicit form. but not traceless.37) Its energy-momentum operator can be derived by standard methods and is equal to θµν = 1 gAB ∂µ φA ∂ν φB − 1 ηµν ∂ρ φA ∂ ρ φB . fµa and φµ can then be written as linear combinations of curls of the independent gauge ﬁelds. At the linearized level. in accord with the earlier counting arguments e given for the gauge ﬁelds.128 Unity from Duality: Gravity. however.34). the currents µ µ Sab . (7. symmetric and traceless (again. let us consider a nonlinear sigma model in ﬂat spacetime with Lagrangian.

43) where the φ-independence of hab can be deduced directly from (7.45) where w is the Weyl weight of the scalar ﬁelds which is equal to w = 1 2 (D − 2). we decompose the target-space coordinates φA into φ and remaining coordinates ϕa .B. this Lagrangian is invariant under local scale transformations characterized by the functions ΛD (x). gaφ ). · · · . 0) and gAφ = χA = (2χ. 4(D − 1) (7.40). These features play an important role when extending to the supersymmetric case. The transformation of gµν is in accord with the vielbein scale transformation written down in Section 7.39) to gravity must lead to a conformally invariant theory of the nonlinear sigma model and gravity. ϕ) = exp[2φ] χ(ϕ). In terms of these new coordinates we have χA = (1. Using the gauge ﬁelds of the conformal group. This result shows that the target space is a cone over a base manifold MB parametrized in terms of the coordinates ϕa with metric hab [106]. de Wit: Supergravity space. e−1 L = 1 gAB ∂µ φA ∂ µ φB − 2 D−2 χ(φ) R . χ can be written as χ = 1 g AB χA χB .44) Indeed. δD gµν = −2ΛD gµν (7. the Lagrangian reads. when the cone is a K¨hler or hyperk¨hler space. 0. δD φA = wΛD χA . (7. Coupling the improved energy-momentum tensor (7. In the supersymmetic context it is important to note that. The relevant Lagrangian reads. where φ is deﬁned by χA ∂ ∂ · = ∂φA ∂φ (7.42) It then follows that χ(φ. We should also point out that the coupling with the Ricci scalar can be understood in the context of the results of the previous section. 2χ (7.46) 2 2 .2. e−1 L = 1 gAB g µν (∂µ φA − w bµ χA ) (∂ν φB − w bν χB ) − 1 w fµ µ χ . In that case U (1) or SU (2) must be associated with the R-symmetry of the superconformal algebra.41) up to an integration constant. To see this. where χ is an undetermined ˆ ˆ function of the coordinates ϕa . From this result one proves directly that the metric takes the form (ds)2 = (dχ)2 + χhab (ϕ) dϕa dϕb . 2 129 (7. Spaces that have such a homothety are cones. the cone must also be invariant under U (1) a a or SU (2).

Theisen. 1 & 2 (NorthHolland/World Scientiﬁc.44). which is still invariant under local dilatations. 1 & 2 (Cambridge Univ. [81]. B. Dine. de Wit and D. Vols. Theor. L¨st and S.47) χ = χ0 . 9 (1997) [hep-th/9709062]. Press. R.43) the Lagrangian then acquires the form. for 4-dimensional N = 2 vector multiplets and hypermultiplets. I & II (Cambridge Univ. The ﬁrst 15-years of superstring theory. 1987). 4(D − 1) (7. Green. J. Kiritsis. coupled to (nonconformal) gravity.Z. 1989). von Gehlen and V. AIP Conf. References [1] P. Supergravity – The basics and beyond. Flume. The constant χ0 appears as an overall constant and is inversely proportional to Newton’s constant in D spacetime dimensions. see. Gauge Theory and Strings As one can easily verify from the transformation rules (7. in Supersymmetry. Freedman. the metric hab should be negative deﬁnite and χ0 must be positive. E. van Nieuwenhuizen. Sezgin.g. u 1989). 1985). 1985). Rep. This example thus demonstrates the relation between improvement terms in the currents and constraints on the gauge ﬁelds. NATO ASI B 125 135. It is possible to also employ a gauge condition for the dilatations. G.48) This Lagrangian describes a nonlinear sigma model with the base manifold MB of the cone as a target space. Vols. we refer to [67. Observe that in order to obtain positive kinetic terms. e. conformal boosts and spacetime diﬀeomorphisms. For a collection of reprints. Supergravities in diverse dimensions.H. 68 (1981) 189. String . An obvious one amounts to putting χ equal to a constant χ0 . e−1 L ∝ 1 hab ∂µ ϕa ∂ µ ϕb − 2 D−2 R. Dietz. [2] M. for example. Salam and E. see. Proc. 419 (1998) 265 [hep-th/9708130]. Vols. M. Witten. Press. Upon using the second constraint (7. For an introduction to the 4-dimensional N = 1 multiplet calculus. 1 & 2 (North-Holland/World Scientiﬁc. J. the Lagrangian becomes equal to (7. Ferrara. The above example forms an important ingredient in the so-called superconformal multiplet calculus that has been used extensively in the construction of nonmaximal supergravity couplings. 1989). Phys. Schwarz. S. (7. D.25). Superstring theory. Schwarz and E. Rittenberg (Plenum. String Theory. Lectures on string theory (Springer.130 Unity from Duality: Gravity. with the dimension D−2 of [mass] . Supersymmetry. 1998). J. Phys.31) for the gauge ﬁeld fµ a associated with the conformal boosts and setting bµ = 0 as a gauge condition for the conformal boosts. Vols. Superstrings. this Lagrangian is invariant under local dilatations. 107]. 1 & 2 (World Scientiﬁc. Leuven Notes Math. A. Polchinski. Substituting the metric (7. There is an extensive literature on this. edited by K. Vols.

Picco and P. c C. D 30 (1984) 325. edited by Dj. Phys. Mod.A. Roˇek (Cambridge Univ. P. de Roo and B. A 2 (1987) 273. edited by S. M.H. Nucl. Louis. Hull and T.H. B 322 (1989) 301. Four lectures on M-theory. Schwarz. Savoy and B.H. L. Howe and P. L. Branes and Dualities. M. 1999).C. C. Lett. Kazakov. Phys. Rev. Bergshoeﬀ. M.B. L¨ st. M. Scherk and J. Maximal supergravity.A. Zumino. Olive. Nucl. Tollst´n and H. p. West. Supersymmetry and dualities in various dimensions.and two-particle states. Phys. J. 318 (1999) 113 [hep-th/9809039]. Lett. Phys. B. Lett.) 55 (1997) 145. Phys. B 122 (1983) 143. 1988). edited by L. B 122 (1977) 253. Phys. Taylor (Cambridge Univ. Nucl. Strathdee. Ferrara and C. Douglas.B. West. Dragoviˇ and D.C. 79 (1981) 1. Baulieu. Representations of extended supersymmetry on one. Nucl. Suppl. U-Duality and central charges in various dimensions revisited [hep-th/9612105]. M. Lett. B 121 (1977) 77. 1999). 2 (1985) 293. Green and J. Nahm. H. B. . C. West. Ferrara and C. Dine. P.M. Di Francesco. Phys. M. Kounnas. Nucl. Phys. P. B 126 (1983) 301. Savoy. in Strings. Phys. D’Auria and S. de Wit and J. edited by S. Phys. Van Proeyen.A. Townsend. edited by L.K. NATO ASI C520. Nucl. Obers and B. Huq and M. Hull. Ortin. 33 [hep-th/9801132]. J. F. I. Schwarz and P. B. V. Int. Douglas. p. Phys. Ferrara. J.G. Nucl.K. Ferrara. B 466 (1999) 144 u [hep-th/9908169]. Phys. in Frontiers in Particle ˇ c Physics ’83. Bergshoeﬀ. Bili´. E. Schwarz. Cremmer. A. Phys. Nucl. Ferrara and J.N.W. N. E. N. J. Nicolai. B 226 (1983) 269. p. Kazakov. N = (4. Giani and M. 152. Press. in Strings. B 100 (1981) 393. P. Nicolai and D. p. Townsend.G. Rep. M-theory from its superalgebra. B 392 (1993) 3 e [hep-th/9208074]. Popoviˇ (World c c c Scientiﬁc. 1984). V. JHEP 0006 (2000) 019 [hep-th/0004086]. Phys. NATO ASI C520. Phys. Rep. in Superspace & Supergravity. Trieste [hep-th/9612121]. F. Phys. Di Francesco. Leigh and J. B. B 135 (1978) 149. B 451 (1995) 547 [hep-th/9504081]. Cremmer. Schwarz. Nucl. 267. M. R. de Wit. Hawking and M. B 76 (1978) 409. 47. W. J. de Wit.A. A. Scherk. P. Phys. B (Proc. Superstring construction (North-Holland. M. Picco and P. Lett. Sijaˇki. S. B 583 (2000) 237 [hep-th/0004195]. in Supergravity ’81. Class. Nucl. S. Slansky. Lett. Schellekens. B 238 (1984) 181. M.C. R. Phys. 141 [hep-th/9712004]. Dai. Nucl. 1989). de Wit: Supergravity 131 [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] Theory in four dimensions (North-Holland. Campbell and P. S. B. Abou-Zeid. Lett.C. Namazie. Quantum Grav. 2) Chiral supergravity in six dimensions and solvable Lie algebras [hep-th/9711048]. R. Phys. A 4 (1989) 2073. Tools for supersymmetry [hep-th/9910030]. Phys. Windey (Kluwer. 1982). Andrianopoli. Pioline.H. Gliozzi. R. Julia and J. p. B 139 (1984) 283. Press. Branes and Dualities. P. lectures given at the 1996 ICTP Summer School in High Energy Physics and Cosmology. Supergravities in 5 dimensions. Mod. Huet and N. Baulieu. Nucl. Pernici. de Wit. J. B 182 (1981) 173. A. Phys. Seiberg. Brink.A. Windey (Kluwer. Ferrara. B. J. de Wit. S. Scherk and D. D’Auria. Polchinski. E.G. 1981). Phys. Phys. E. B 243 (1984) 112.

M. de Wit. Phys. Bautier. van Nieuwenhuizen. Phys. Lett. B 85 (1979) 219. Phys. J. Cremmer. Phys. edited by K. S.T. G¨ven. Modern Kaluza-Klein Theories (Addison-Wesley. Proc. c [51] B. Phys. Proc. [44] M. D 18 (1978) 3630.K. B 86 (1979) 161. Quantum Grav. 1987).J.J.C. 1981).W. Seminara. Zumino.A. de Wit and D. in Supersymmetry and Supergravity ’84. u [46] For a review see.Z. Townsend and P. [33] S. Vanderseypen and A. P. B 129 (1977) 125. in Supergravity. Int. Maldacena. de Wit. Mod. A 13 (1980) 1643. Lett. B (Proc. Ferrara.A. M. B 129 (1977) 39. B. 331. Mod. Phys. edited by M. and references quoted therein. L¨ . Rev. 2 (1998) 231 [hep-th/9711200]. 1980). Duﬀ and K.K. Rev. A. High energy physics and cosmology.A. Nicolai. Phys.J. F. in Uniﬁcation of fundamental particle interactions. de Wit. edited by S. Fronsdal. Phys. Phys. [50] E. Nilsson and C. Fang and C. p. Olive.H.132 Unity from Duality: Gravity. [35] K.J. Quantum Grav.K. Warner. M. Liu and J. Curtright. p. [47] B. Stelle. Introduction to supergravity. de Wit and A. [42] P. edited by P. Duﬀ. [40] B. Freedman. [52] Y. Lett. Nucl. in Trieste 1996. Phys. B 69 (1977) 481. Aragone and S. Phys. Lett. [29] M. [43] D. Page. West. B. u 259 (1995) 213 [hep-th/9412184]. [39] M. 1984). Thirty Years of Supersymmetry. Ferrara. Lu [hep-th/0010292]. [49] B.X. Class. van Nieuwenhuizen (Plenum. B 145 (1984) 368. Phys. Schwinger. B 406 (1997) 49. Ellis and P. Lett. Strings 2000. [38] C. Phys. Phys. Rev. Phys. 1979). Lett. Fayet and P. B. [36] C. D 15 (1977) 2802. Nucl. [45] R. de Wit. 1970). B 272 (1986) 598. see. Chodos and P. B 124 (1983) 45. [48] For a reprint collection and a review of Kaluza-Klein theory.G. [28] J. [34] B. Townsend. Coquereaux. D 28 (1983) 2976. A. Rudaz and M. L. Freund. sources and ﬁelds (Addison Wesley.S. de Wit. Nicolai. F. Roˇek (Cambridge University Press. in Superspace & Supergravity. Phys. Duﬀ. 137. Aragone and S. D 21 (1980) 358. Kaku. N=8 Supergravity. Rep. Lett. edited by B.W. Rev. Phys. Phys. T. B 67 (1977) 439.) 101 (2001) 154.Z. de Wit and H. Phys. Ferrara. Phys. Adv.P. B. for example. K. Lett. F. van Nieuwenhuizen (World Scientiﬁc. P. de Wit. van Nieuwenhuizen and D. S. Lett. Zwartkruis. 4 (1987) L59. Phys. Hawking and M.O. Stelle.W. Rep. Van Proeyen. T Appelquist. Vasiliev. 287 [hep-th/9701088]. p. J. 18 (2001) 3095 [hep-th/0011239]. [31] P. B 115 (1982) 389. A 16 (2001) 1002. Phys. Suppl. Julia. de Wit and H. D 5 (1996) 763 [hep-th/9611024]. Int. van Nieuwenhuizen and B. Lectures on supergravity p-branes. P. [37] C. B 400 (1993) 463 [hep-th/9210068]. Lett. . J.S. Class. J. for instance. Phys. Biran. Tanii. J. Deser. Romans and N.J. Henneaux and D. Duﬀ. u [30] B. Theor. Particles. Math. Shifman [hep-th/0103086]. Phys.E. Freedman (North-Holland.R. Phys. p. Englert. edited by S. [27] S. Phys. 49. Berends. Nucl. Pope. J. B 253 (1991) 113. Khuri and J. Chamseddine and P. Townsend and P. R. B 276 (1992) 49. [32] S. Spin 2 matter-graviton coupling problems. “Group disintegrations”. Fronsdal. van Holten. 38 (1977) 1433.N. J. J. Rev. 130 (1986) 1. D 18 (1978) 3264. Gauge Theory and Strings [26] R. Lett. van Nieuwenhuizen.N. M. Deser and B. G¨naydin. Phys. [41] B. Rev. Deser. Deser. Nucl. Nucl.

B. P. Phys. D’Auria and P. Phys.R. B 350 (1995) 184 [hep-th/9501068]. Townsend. 2 (1985) 343. Duality and spontaneously broken supergravity in ﬂat backgrounds [hep-th/0204145]. Ferrara and M. B 136 (1984) 38.H. 79 [hep-th/9908005]. van Holten and A. Nicolai.B. B. Herger. Kazakov. Lett. P. Ferrara and L. Manton. Lett. Gaillard and B. B. Rev. Pilch and P.S. Phys. Lett. Phys. Phys. Phys. Bergshoeﬀ.N. A 4 (1989) 2475. Phys.-W.Andreanopoli. R. Lett. Phys. P.K. Hawking and G. JHEP 0104 (2001) 022 [hep-th/0103032]. Ann. L. Trigiante (to appear). D 18 (1978) 3565. J. J. Phys. New gauged N = 8. . G. Lett. Press. R. Nucl. Phys. Girardello. K. P. S. P. Isham and D. B 373 (1996) 68 [hep-th/9512062]. 1983). D 24 (1981) 3065. Polchinski and E. Nucl. B 189 (1987) 75. K. Phys. M. Lett. B. de Wit and A. D = 4 supergravities [hep-th/0204156]. 18 e (2001) 395 [hep-th/0009048]. The large scale structure of space-time (Cambridge Univ. Lett. Suppl. Nucl. B (Proc. Van Proeyen.A. Lect. Avis. Phys. S. 1973). Rev. Taylor and P. J. de Wit. Ferrara. H. J. Chapline and N.H. Lie algebras. H. H. Cecotti. Phys.J. Phys. Towards Quantum Gravity. B 148 (1984) 297.G. Antoniadis. B 208 (1982) 323. Van Nieuwenhuizen (World Scientiﬁc. E. L. Zumino. de Wit and H. Phys. Bergshoeﬀ. Multiplet calculus. Samtleben. Lett. Witten. F. Phys. Phys. Nucl. Pilch and A. edited by S. Fr´ and L. Taylor. B. de Wit. B 438 (1995) 109 [hep-th/9410167]. Lled´. B 245 (1984) 89.K. K. E. Picco and P. Gilmore. B. S. de Wit. Townsend. B 367 (1996) 97 [hep-th/9510086]. Aspinwall. Math.R.) 46 (1996) 30 [hep-th/9508154]. M. B 193 (1981) 221. Townsend.K. 541 (2000) 79. 1999) p. Castellani. Samtleben and M. de Wit and I. edited by L. Phys. Pernici. 179 [hep-th/9706211]. Windey (Kluwer. B 142 (1984) 39. I. C. 185 (1988) 330. B 195 (1982) 97. E. Rev. Sezgin and P. Nicolai and H. 25 (1984) 3455. Douglas. C. J. Notes Phys. B 120 (1983) 105. H. Chamseddine. M. M. Baulieu. L. Nucl. Nucl. de Wit: Supergravity [53] [54] [55] [56] [57] [58] [59] 133 [60] [61] [62] [63] [64] [65] [66] [67] [68] [69] [70] [71] [72] [73] [74] [75] [76] [77] [78] [79] [80] [81] E.S. van Nieuwenhuizen. in Supersymmetry and Supergravity ’82. Di Francesco. Class. 1991). S. de Wit and P. Rev. Phys. Phys. de Roo. Nucl. S. NATO ASI C520. Phys. also in Polanica 1999. Partouche and T. Witten. Nucl. A 15 (1982) 3763.M. B 108 (1982) 285. J.M.W. V. 1974). JHEP 0207 (2002) o 010 [hep-th/0203206]. Supergravity and superstrings I (World e Scientiﬁc. Phys. Andreanopoli. Townsend. Julia. p. A. Phys. D’Auria. Qunatum Grav. Nucl. Phys. Cremmer and B. Lett. B 80 (1978) 48. Hull and P. Lett. B 217 (1983) 143. Samtleben. E. Lett. Schellekens.K. Fr´. Phys. Storey. J.F. Mod. C. Schwarz. E. M. D 30 (1984) 760. de Roo and B. Phys. Pilch and P. Lie groups. Nucl. B 460 (1996) 525 [hep-th/9510169]. de Wit. B 443 (1995) 85 [hep-th/9503124]. van Nieuwenhuizen. 86 (2001) 1686 [hep-th/0010076].J. Phys. Bergshoeﬀ. M. R. van Nieuwenhuizen. B 158 (1979) 189.K. B 143 (1984) 103. Nicolai and H. Phys. Hull. Lett.A. B. Van Proeyen. and some of their applications (Wiley Interscience. Ellis. Phys. Gualtieri. Cordaro. Class. Int. Quantum Grav.

Nucl. Conformal higher spin theory [hep-th/0207212].N. M. Haugen. Gubser. edited by B. de Wit. Minic and M.K. Lecture o u Notes Phys. Heidenreich. Maldacena. Phys. Townsend and P. Phys.-W. Polyakov. Freedman. Oscillator like unitary representations of noncompact groups and suu pergroups and extended supergravity theories. Sezgin and P. [89] M.G. J. Quantum Grav. J. Phys. Skenderis and S. Phys. [83] C. [91] W. in Supergravity ’81. Nucl. Phys. Colloq. Townsend and P. 53 (1969) 174. P. [101] S.P.R. de Wit and J. [hep-th/9712239]. 91 (1982) 159.S. Massless higher spins and holography [hep-th/0205131]. Phys. Kraus and A. Rev. de Wit. Nucl. P. Math. Warner. [88] M. van Holten. Gubser.J.N.-W. B 110 (1982) 461. van Nieuwenhuizen and N. Freedman and H. de Wit. [86] P. Fronsdal and R. B 237 (1984) 342. 1990). Math.Z. 37 (1965) 221. Oz. P. In¨n¨. Sezgin (World Scientiﬁc. Solodukhin. G¨naydin and D. Gauge Theory and Strings [82] P. Zagerman. Theor. B. Salam. Phys. B 443 (1998) 138 [hep-th/9809158]. Phys. G¨naydin D. van Nieuwenhuizen (World Scientiﬁc. edited by M. Nucl. van Holten and A. [93] S. B 155 (1979) 530. Witten. Rev. [85] H. u G¨naydin. Ferrara and C. Phys. Phys. Taylor (Cambridge Univ. Nucl. de Haro. van Nieuwenhuizen. Pope and E. Nucl.M. Class. Duﬀ. Phys. p. Phys. Ann.B. Minic. Math. on Group Theoretical Methods in Physics (Istanbul. 144 (1982) 249. 1984). [95] A. B 255 (1985) u 63. P. Singleton and doubleton supermultiplets of space-time supergroups u and inﬁnite spin superalgebras. Ann. Class. [92] M. Sundell. Mod. 180 (Springer. Marcus. Commun. Rep. C. Phys. 217 (2001) 595 [hep-th/0002230]. Ferrara and J. [87] M. Dirac.Z. I. C. 2 (1998) 253 [hep-th/9802150]. Math. Rychenkova. in Supersymmetry and Supergravity ’84. [106] G. G¨naydin and C. Phys. [98] E. D 12 (1975) 3810. Rev. Nicolai. Phys. Commun. Fronsdal. Mack and A. B 428 (1998) 105 [hep-th/9802109]. Aharony.M. B 523 (1998) 145 [hep-th/9802047]. D 19 (1979) 3166. G¨naydin u u and N. [97] S. Phys. 1983). S. Adv. Conformal invariance in extended supergravity. Breitenlohner and D. Kaku. [84] D.A. Fayet. Nicolai.S. H. Phys. D 59 (1999) 046003 [hep-th/9805171]. Rev. 4 (1963) 901. 2 (1985) L11. Phys. Klebanov and A. Press.W. Y. D 17 (1978) 3179. van Nieuwenhuizen. 1982). C. Segal. Fronsdal. . G¨naydin. Serdaroglu and E. [96] J. Brown and M. D 10 (1974) 589. Representations of supersymmetry in anti-de Sitter space.134 Unity from Duality: Gravity. 104 (1986) 207. Lett. [105] B. 323 (2000) 183 [hep-th/9905111]. [102] G. [104] M. Nucl. in Supermembranes and physics of 2+1 dimensions. Fronsdal. Phys. M. Saclioglu. [90] M. u M. in Int. B 272 (1986) 99. Rev. Quantum Grav. Gibbons and P. G¨naydin. D 12 (1975) 3819. Lawrence. Ooguri. J. Van Proeyen.K. Commun. Rev. B 534 (1998) 96 u [hep-th/9806042]. G¨naydin and N. B 167 (1980) 186. Rev.D. Phys. Warner. [99] V. 15 (1998) 2153 [94] E. Phys. Lett.P. P. [103] B. Phys. 442. 1982). [100] O. Math. Phys. Phys. edited by S. Lett. edited by M. K.M. Phys. Balasubramanian. Henneaux. P.

37 [hep-th/9808160]. 9 (1992) 27. Phys. Roˇek and S. Nucl. p. Rigid N = 2 superconformal hypermultiplets. Vandoren. B. . B. de Wit.B. Galicki. B 568 (2000) 475 [hep-th/9909228]. Class. de Wit. Quantum Grav. JHEP c 0102 (2001) 039 [hep-th/0101161]. Vandoren. M. in Dubna 1979. Supersymmetries and Quantum Symmetries. de Wit: Supergravity 135 [107] K. Kleijn and S. B.

Jerusalem. Hebrew University.LECTURE 2 SUPERSYMMETRIC GAUGE THEORIES E. RABINOVICI Racah Institute. Israel .

. . . . . . . 2. . . . . . . . . . . . . . . . . . . . .10 Superconformal invariance in d = 4 . . . . . . . . . . . . . . . . . . . . . . . . . . .7 NF = NC + 1 . . . . . . . . . . . . . .4 Integrating in . . . . . . 2. 3 Review of supersymmetric models 3. 5. . .4 F-terms . . . 5. . . 3. . . . . . . . . . . . . . . . . . . . . .2 A nonrenormalisation theorem . . . . . . . 2. . . . . . . . . . . . . . . . . 5. . . . .3 Quantum moduli space for 0 < NF < NC 5. . . . . .2 Quantum moduli spaces . . . .Contents 1 Introduction 2 Supersymmetric quantum mechanics 2. . 5. . . . . . . . . . . . . . . . . . 2. 3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4 Geometric meaning of the Witten index . . . . . . . a 3. . . . . .3 K¨hler potentials . . . . . . . . . . . . . . . . . . . . . . . 2. . . . . . . . . . . . . . . . . . . . . . . . 3. . . . . . .7 Supersymmetry breaking . . . . . . 3. .6 The eﬀective potential . . . . . . . . . 3. . . . . . . . . .1 Kinematics . . . . . . . . . . . . . . . .7 Superconformal quantum mechanics . . . . . . . .3 A two variable realization and ﬂat potentials 2. . . . . . . . . . . .5 Quantum moduli space for NF ≥ NC . . . . . .8 Higgs and conﬁnement phases . . . . . . . . . . . . . . . . . . . . . . . . . .6 NF = NC . . . . . . . . . . . . . . . . . . 5. . . . . . . . 5. . . . .2 Superspace and chiral ﬁelds . . . . . . . . . . . . . . .5 Landau levels and SUSY QM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5. . . . . . . . . 3. 6 Comments on vacuum energies in scale invariant theories . . . . . . . . . . . . . . . . . . . . . . .8 Supersymmetric gauge theories 4 Phases of gauge theories 5 Supersymmetric gauge theories/super QCD 5. .9 Infra-red duality .1 The classical moduli space . . . . . . . . . . . .5 Global symmetries . 5. . . . . . . . . . . . . . . . . . . . . . . .6 Conformal quantum mechanics . . . . . . . . . . . . . . . . 141 143 151 152 154 158 159 161 164 165 165 167 169 170 170 172 172 173 179 180 182 185 185 186 190 190 191 192 194 202 207 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 Symmetry and symmetry breaking . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . 7. . 7. . 7. . . .5 An eﬀective D = 4. . . . . . . . . . . . . . . . . . . . . 8 Final remarks . . . . . . . . .8 Concluding remarks . 210 210 211 213 216 226 229 233 235 237 . . . . . . .4 Eﬀective D = 4 dimensional systems with N = 2 supersymmetry 7. 7. . . . U (NC ) gauge theory with matter . N = 1. .3 The eﬀective ﬁeld theory on branes . . . . . . . . . . . . . . . . . . . . . .1 Branes in string theory .2 Branes in IIA and IIB string theories . . . . . . . . . . . . . . . 7. . . . . .7 Obtaining the dual ﬁeld theory . . . . . . . . . . . .6 More pieces of information . . . . .7 Supersymmetric gauge theories and string theory 7. . . . 7. . . .

The weak interactions are in the Higgs phase. the force in nature that has yet to be tamed by theoretical physics.the American-Israeli Bi-National Science foundation. little did they know that the constrained interactions [2] will give rise to a multitude of vacua. This is done by ﬁrst describing various supersymmetric quantum mechanics models. the GermanIsraeli Bi-National science foundation.SUPERSYMMETRIC GAUGE THEORIES D. What is then the motivation to introduce and study supersymmetry? Here is a list of reasons that have motivated people over the years. the phase structure of supersymmetric systems with and without gauge particles. 1 Racah Institute. A discussion of the relation of conformal symmetry to a vanishing vacuum energy (cosmological constant) is included. Berman1 and E. • In their seminal paper Golfand and Likhtman [1] introduced supersymmetry in order to constrain the possible forms of interaction. This work is supported by BSF. The electromagnetic interaction is in the Coulomb phase and the colour interactions are in the conﬁning phase. • The local version of supersymmetry contains automatically gravity [3. Israel. the Israel academy of sciences and humanities-Centers of excellence program. 1 Introduction At least three phases of gauge theory are manifest in nature. Rabinovici1 Abstract We introduce simple and more advanced concepts that have played a key role in the development of supersymmetric systems.S. superconformal theories and infrared duality in both ﬁeld theory and string theory. Jerusalem. c EDP Sciences. Springer-Verlag 2002 . 4]. Hebrew University. the european RTN network HPRN-CT-2000-00122 and DOE grant DE-FG02-90ER40560. Topics covered include the basic construction of supersymmetric ﬁeld theories. None of these phases exhibits supersymmetry explicitly.

142 Unity from Duality: Gravity. In section ﬁve we describe the phenomena of infra-red duality in supersymmetric gauge theories. in order for string theories to make sense the number of Bosonic degrees of freedom can diﬀer from those of the Fermionic degrees of freedom by at most those of a d = 2 ﬁeld theory [11]. In the presence of supersymmetry constraints the problems can be solved in interesting cases. This duality is described from both a ﬁeld theoretical and string theoretical point of view. This section includes also a discussion of conformal ﬁeld theories and some properties in dimension greater than two including four. 12–18]. This problem is much softer in supersymmetric theories [5–7]. They include the understanding of conﬁnement. 15]. we refer the reader to them for a more complete exposition [3. The ideas include: supersymmetry and its spontaneous breaking. In particular. index theorems as a tool to answer physical questions. the impact of conformal symmetries. chiral symmetry breaking and the emergence of massless fermions to mention just a few. called the Hierarchy problem. In particular the idea of monopole condensation as the responsible mechanism for conﬁnement becomes tractable. There are many excellent reviews on the subject. The construction and analysis of supersymmetric systems requires the use of many diﬀerent concepts and ideas in ﬁeld theory. In section four. the material in this section is elaborated in similar or greater detail in [12. the vacuum energy of these theories is discussed in relation to the cosmological constant problem. Section three will contain a review of the methods usedto construct supersymmetric ﬁeld theories and supersymmetric gauge theories. and treating theories with no ground states. Such models serve as analytical laboratories for physical ideas. In fact. In section two we will introduce several of these ideas by using a simple quantum mechanical context. In the process we will discuss connections between string theory in the presence of branes and gauge theories. we review the intricate phase structure of classical and quantum supersymmetric gauge theories. Gauge Theory and Strings • The weak interaction scale and the Plank scale are widely separated for theoreticians this could be a problem. . It may occasionally seem that one sees only the trees and not the forest however note that many of the trees are central problems in ﬁeld theory for years. • Perturbative string theory seams to need supersymmetry in order to be stable. • In various supersymmetric models the gauge couplings unify sensibly [8–10]. • Strong coupling problems are generically intractable in ﬁeld theory. 4.

ψβ ] 2 α (2. the basic anti-commutation relations that deﬁne the supersymmetric algebra are: {Qi . Hamiltonian is then given by: ∂2W 1 ∂W 2 1 p2 + Bαβ 1L×L − H = {Q.. j = 1.. xn ) is a general function of the n bosonic variables.2) The W (x. Examples of such topics are: large N gauge theories [19]. 2 Supersymmetric quantum mechanics The ideas in this section were introduced in [24. Berman and E.D. N denotes the number of supersymmetries. Along the way we will introduce some useful tools such as the Witten index.S. .g. (2. [20]. non-supersymmetric deformations of the models described e. Q+ are called the supercharges. supersymmetry on the lattice [22] as well as many others. The Hamiltonian is the only member of the Poincare group in that case. how is SUSY broken spontaneously. The questions we wish to examine are: what is supersymmetry. supersymmetric matrix models [21]. First we will examine the quantum mechanical realization of the supersymmetry algebra so as to introduce various ideas that will later carry over to ﬁeld theory.N (2.1) H is the Hamiltonian and Q. Rabinovici: Supersymmetric Gauge Theories 143 Many topics in the study of supersymmetric theories which are as worthy have not been covered for lack of time or because they were covered by other lectures. Now we begin with a one variable realization of N = 1 SUSY. 25]. A rather general N = 1 realization with n bosonic and n Fermionic degrees of freedom is given by: n Q= α=1 + ψα −pα + i ∂W ∂xα .3) α 2 2 ∂xα ∂xα ∂xβ α αβ and B is: Bαβ = 1 + [ψ . Later we will present a two variable realization where we can introduce the notion of a ﬂat direction. and what are nonrenormalisation theorems. Q+ } = 2Hδij j i.4) . Q+ } = (2.. Thus. Quantum mechanics is a one dimensional ﬁeld theory.

5) The dimension of the Fermionic Hilbert space is L = 2n . (2. (2. a+ . {ψα . we ﬁrst will recall some basic facts about the Bosonic harmonic oscillator. N is constructed out of the creation and annihilation operators: N = a+ a.12) The spectrum of this operator is given by the nonnegative integers. Gauge Theory and Strings and the ψ variables obey the following anticommutation relations. . px = (mω )− 2 pq . a by: 1 i x = √ (a + a+ ) . a+ ] = 1. q] = [px .144 Unity from Duality: Gravity.11) (2.8) ω is the energy scale in the problem.10) The number operator. H= p2 1 q + mω 2 q 2 2m 2 (2. Before moving on to n = 1 realization of this super algebra.9) (2. the Hamiltonian is now given by: 1 H = ω (p2 + x2 ) 2 x (2. ψβ } = 0 .6) deﬁne The energy scale is extracted by deﬁning dimensionless variables x and px . p = √ (a+ − a) 2 2 and so we obtain the commutation relation: [a. It is useful to deﬁne creation and annihilation operators. (2.13) and the energy spacings are given by the scale ω. x] . {ψα .7) This gives the following commutation relations: −i = [pq . + + + {ψα . ψβ } = δαβ . ψβ } = 0. x= mω 1 2 q. 1 (2. The Hamiltonian in these variables then becomes: H= ω N+ 1 2 (2.

14): (x + ip)φ(x) = 0. This is essentially due the uncertainty relation.18) x2 2 · (2. Berman and E.17) which can now be expressed in x-space. The eigenstates.15) yielding the ground state wave function: φ(x) = c exp − The state |n is given by: (a+ )n |n = √ |0 . 2. That is: x+ d dx φ(x) = 0 (2.D. The ground state. (2. |0 is obtained by solving the equation: by a|0 = 0. Using equations (2.19) So why not obtain a ground state energy of − 1 ω by solving 2 a+ |GS = 0. 2 (2. This completes the solution of the Bosonic harmonic oscillator. Rabinovici: Supersymmetric Gauge Theories 145 The energy of the ground state diﬀers from the classical minimum energy 1 2 ω.14) Note that this equation is a ﬁrst order as opposed to the second order equation that one would have to solve if one attempted to directly solve the Schroedinger equation. |n can be recast in terms of the variable x.20) To ﬁnd a quantum state it is not enough that it be a zero energy solution of the Schr¨dinger equation but it must also be normalizable (by normalizable o we include plane wave normalisability).10.21) (2. Also note that one could have written H as: H = ω aa+ − 1 2 · (2. This is possible only for the ground state. n! (2. This ﬁrst order equation always has a solution. .16) (2. It is given by: φ(x) = c exp 1 2 x .S.

mf 2 (m. giving 2H = and so one may choose gf (m. We will now note some basic facts concerning the validity of a perturbative expansion. x that preserve the commutation relations. such that: [pq . 2m 2 (2.23) (2. Gauge Theory and Strings The above solution is not even plane wave normalizable and hence it is not a quantum state. g) (2. x] and 1 H = h(m. g) 1 n+2 (2. 2 x We will make the following ansatz: q = f (m.28) .24) pq = 1 px . the universe may well be ﬁnite. q] = [px . g) (p2 + xn ). m dependence the same type of rescaling used for the harmonic oscillator (2.146 Unity from Duality: Gravity. In such a circumstance the wave function will be exponentially conﬁned to the edge of the universe and thus irrelevant for bulk physics.27) (2. Consider the Hamiltonian. There may be situations where one will want to study the physics on the boundary. g) = The Hamiltonian becomes: n n 1 H = g 1− n+2 m− n+2 (p2 + q n ). the simple harmonic oscillator is a useful approximation for small excitations above the minimum of a generic potential. g)2 · (2.22) One may wonder whether if one can make a perturbative expansion in small or large g or small or large m. To answer this one needs to ﬁnd out if one can remove the g. Also recall that. g)n xn . f (m. H= p2 1 q + gq n . how do we motivate this restriction? After all.26) 1 mf (m. In such cases the “non-normalizable” states should be kept and may play an important role.7). g)x . 2 q (2. From now on we will only accept plane wave normalizable states and nothing “worse”. Is it possible to deﬁne a new set of dimensionless canonical variables px . Physically.25) p2 x + gf (m.

By analogy. which we will return to later. After reviewing the Bosonic harmonic oscillator. a+ } = 1 . [a+ .35) (2. F F [aF .30) This demonstrates that a+ . {aF . The states are given by |0 . F The spectrum of NF is 0. (2. perturbation theory is valid. aF } = {a+ . This is essentially a manifestation of the Pauli exclusion principle. F f F (2. 1. F where ωF is the Fermionic oscillator frequency and. σ− = We then identify. F (2. Hamiltonians of the form p2 + xn can’t be analyzed perturbatively. we have HB = 1 1 ωB {aB . {aF . H] = −a+ . H = ωF (a+ aF + const. (This does not apply to the special case of n = −2. we review the harmonic Fermionic Oscillator. They may not serve as perturbation parameters.) . aF are creation and annihilation operators. aF = σ− . aF ] = ωF NF − F 2 2 · (2. we choose HF to be.29) The Hamiltonian (that does not have a classical analogue) is taken to be. a+ = σ+ . σ+ = (σ1 + iσ2 ) · 2 2 (2.32) This algebra can be realized using the Pauli matrices {σ i } as follows. Further analysis shows that for Hamiltonians of the form p2 + xn + xm with m < n.) Thus. a+ } = 0. a+ } = ωB NB + B 2 2 · (2. Berman and E. Rabinovici: Supersymmetric Gauge Theories 147 The role of g and m is just to determine the overall energy scale.34) 1 1 (σ1 − iσ2 ) . The commutation relations become anticommutation relations. F Hence the Hamiltonian becomes: H= For the Bosonic case.S.D. HF = 1 1 ωF [a+ . 2 (2.31) (2. H] = aF .37) . Deﬁne.33) 1 This shifts the energy levels by + 2 ω. a+ |0 .36) 1 ωF (σ3 + 1) + c.

46) (2. 0 ) 1(Q|E. |E. Gauge Theory and Strings With this choice. 2 (2. 0 ) = Q(Q|E.148 Unity from Duality: Gravity. NF = 0. 0 denotes a state with energy E and NF = 0. 0 ) = NF (Q|E.41) Q2 = = = 1 2 2 2 σ p + σ2 x2 + σ1 σ2 px + σ2 σ1 xp 2 1 1 2 p + x2 12×2 + iσ3 (px − xp) 2 1 2 1 p + x2 12×2 + σ3 2 2 (2.45) Q anticommutes with σ3 and commutes with H.40) (2. 0 ) E|E.48) . We will label a state by its energy E and its Fermion number.e. 0 .42) (2. i. H(Q|E. Deﬁne 1 Q = √ (σ1 p + σ2 x) 2 1 {Q. the vacuum energy may cancel between the Bosons and Fermions. 0 ) = E(Q|E.44) hence. Q} = Q2 2 (2. We see that the total Hamiltonian becomes: H = ωB NB + 1 2 + ωF NF − 1 2 · (2. (2.38) The Fermionic and Bosonic number operators both commute with the total Hamiltonian.47) (2. 1. Consider the case when ω = ωF = ωB . Thus. En = ω(nB + nF ). here the vacuum energy precisely vanishes.39) H= Q commutes with H. There is now a symmetry except for the E = 0 state.43) (2. (2. H= 1 ω (p2 + x2 12×2 + σ3 ).

S. NF with E = 0 there is a state with equal energy and diﬀerent NF obtained by the application of Q. H = Q2 + Q2 = 1 2 Thus E ≥ 0 and for E > 0 the spectrum is paired.52) (2.56) In the previous case. 0 ) = Q(|E. Qi ] = 0 . σ3 } = 0. 0 however is annihilated by Q and so is not necessarily paired. Q− } = 0.D. Rabinovici: Supersymmetric Gauge Theories This implies Q(|E. Q1 = iσ3 Q2 . The ground state |0. The 2 generalized Hamiltonian is now: 1 2 1 (p + W (x)2 )1 + W (x)σ3 2 2 [H. a simple harmonic oscillator).58) . For any state |E. {Qi . It can be generalized to more complicated cases.55) (2. Note.54) This was a “free” theory. W (x) = 1 x2 .e. Q+ } = 0 . 1 ) = √ E|E. (i.49) (2. (2. 0 149 (2. Q2 } = 0. Berman and E.53) (2.50) which is valid only for E = 0. (2. 2 (2. {Q1 .57) (2. Q− } . with the simple harmonic oscillator. 1 √ E|E. We introduce new supercharges that depend on a potential W (x) as follows: Q1 Q2 = = 1 √ (σ1 p + σ2 W (x)) 2 1 √ (σ2 p − σ1 W (x)). 2 1 1 Q2 = √ (σ2 p − σ1 x) = √ σ2 (p + iσ3 x) 2 2 and. {Q− . [H. that there is a Q2 such that also H = Q2 . σ3 ] = 0 . We can then deﬁne: 1 Q± = √ (Q1 ± iQ2 ) 2 1 {Q+ .51) H= (2. 2 {Q+ .

One can now solve the ﬁrst order equation to ﬁnd the zero energy states. Consider the E = 0 case. So Hφ = 0 ⇔ Qφ = 0. In ﬁeld theory the energy gap between he ﬁrst excited state and the ground state is the particle mass.67) (2. (2. It is necessary to solve two 2nd order equations: p2 + (W )2 + W 0 0 p2 + (W )2 − W φ1 (x) φ0 (x) = 2E φ1 (x) · (2.63) (2. (2.150 Unity from Duality: Gravity.68) .66) (2. NF = 0) = 1 and n(E = 0. In general one may have any number of zero energy states in each NF sector. thus the mass of a free Boson equals that of a free Fermion. Qφ = 0 1 √ (σ1 p + σ2 W (x))φ = 0.60) Where we have used that Q is Hermitian. can be found (if they exist) for any W (x) by using the following: Hφ = Eφ . Proof: 0 = 0|H|0 = 0|QQ|0 = (||Q|0 ||) ⇒ Q|0 = 0.64) These can be integrated to give the following solutions: φ1 (x) φ0 (x) = = φ1 (0) exp(−W (0)) exp(W (x)) φ0 (0) exp(W (0)) exp(−W (x)). n(E = 0.65) (2. H = Q2 . NF = 1) = 0 2 where n denotes the number of states of given E and NF .62) (2.61) (2. The energy gap in the Bosonic sector (NF = 0) matches the energy gap in the Fermionic (NF = 1) sector. The full spectrum can’t be solved for general potential W (x).59) φ0 (x) The zero energy solutions on the other hand. (2. 2 This leads to two independent ﬁrst order diﬀerential equations: − d + W (x) φ1 (x) dx d + W (x) φ0 (x) dx = = 0 0. Gauge Theory and Strings There is an analogue in ﬁeld theory. For W (x) = 1 x2 .

The relevant question is whether Q|G. g and the symmetry is spontaneously broken if g|G. denote the energy ground state of the system. = 0 and Q|G. Berman and E. It will appear again in a variety of contexts. = 0 iﬀ EG.S.S. = 0. This corresponds to a Bosonic ground state. Returning to our model.S.S. if S|G. it has generators.S. If the ground state is not invariant under S. [H. In the case of a continuous symmetry. such a symmetry breaking may result in massless particles (called Goldstone Bosons). as |x| → ∞ then both solutions are possible. S] = 0. Q|G. If on the other hand W (x) → −∞ as |x| → ∞ then φ1 (x) is a normalizable solution and one must set φ0 to zero and obtain a Fermionic ground state. then it is said that the symmetry is spontaneously broken.S.S. if Q|0 = 0 then there is a supersymmetric ground state and spontaneous supersymmetry breaking does not occur. For W (x) → const. For the N previously considered potential. Rabinovici: Supersymmetric Gauge Theories 151 A zero energy solution always exists. one sees that if N is even then there is no spontaneous symmetry breaking since there are zero energy normalizable solutions and conversely if N is odd then there is spontaneous symmetry breaking. Assume S is a symmetry of the Hamiltonian. When there is no spontaneous supersymmetry breaking Fermions and Bosons have the same mass. If supersymmetry is spontaneously broken then the ground state will not have zero energy. as it occurs in the Zeeman eﬀect where one introduces a magnetic ﬁeld in the action that does not possess all the symmetries of the original action. There is at N most one unpaired solution at E = 0. Q is the generator of a continuous symmetry (supersymmetry). This implies there will be at least two ground states with identical .S. There is no E = 0 solution if N is odd. that is. For W (x) = n=0 an xn there is an E = 0 solution if N is even.D. = 0. Previously it was shown that any state with nonzero energy must be paired with at least one other state (its supersymmetric partner). Spontaneous symmetry breaking plays an important role in the standard model and in a variety of statistical mechanical systems.S. The consequences on the spectrum are as follows. The fact that nE=0 depends on such global information is very useful.1 Symmetry and symmetry breaking One can break a symmetry explicitly. = 0.S = |G. Let |G. W (x) = i ai xi . Only the leading asymptotic behavior matters. 2. This occurs when the ground state does not posses the symmetry of the action. When the symmetry is continuous. Is it a physically acceptable solution? If W (x) → ∞ as |x| → ∞ then φ0 (x) is a normalizable solution and one must set φ1 (x) = 0. = 0 iﬀ EG. The symmetry may be broken spontaneously.

H→ 1 (W (x))2 ≥ 0. For example. Gauge Theory and Strings No Susy E Susy Susy mB mF E=0 mB mF NF=0 NF=1 NF=0 NF=1 NF =0 NF =1 mB=mF >0 mB>mF =0 Fig. It arises from the breaking of a Fermionic symmetry just as the Goldstone Boson arises from breaking a Bosonic symmetry. This the same as the exact result calculated previously by solving the Schroedinger equation! So how .152 Unity from Duality: Gravity. The Goldstone Boson and the Goldstino have special low energy couplings. The classical limit is taken by → 0. For any choice of ground state. Classically therefore the ground state has zero energy. 1.69) correc- First we will do a classical analysis and then we will examine the tions to the classical result. 2 (2. Spectra. The zero mass Fermion is called the Goldstino. energies and one must make a choice for the vacuum.70) For N even W (x) has at least one zero where W (x0 ) = 0. taking the ground state to be the state with Fermion number zero implies that there is a Fermionic state with the same energy as the ground state thus the Fermion mass will now be zero while the Boson will remain with mass ∆. there is a zero energy gap between that ground state and the other possible ground state with Fermion number diﬀerent by one. Hamiltonian is H= 1 2 − 2 The ∂2 1 + W (x)2 1 + σ3 W (x). After spontaneous symmetry breaking the Boson and Fermion masses are no longer equal. 2.2 A nonrenormalisation theorem Consider now the system from a perturbative perspective. ∂x2 2 (2.

2 (2.73) Thus the Bosonic correction is canceled by the Fermionic correction and the classical result is exact.72) In a theory with Fermions though. In this case the classical result is not exact. the potential is: |W (x) |2 = (x2 + a)2 . there is no ground state with E = 0. However.S. (2.74) This also has E = 0 classical solutions and the leading term in W (x) is odd thus there are no exact quantum E = 0 states. (2. There are nonperturbative eﬀects that provide corrections. Perturbatively. the energies of these two states are equal. Rabinovici: Supersymmetric Gauge Theories 153 do classical results become exact? First let us note that in the Bosonic case. N is odd and the supersymmetry is broken. Calculating the perturbative energy of the ground state for a supersymmetric theory we have: SUSY Epert = 1 ((W (x0 ))2 + W (x0 ) − W (x0 )). 2 (2. Berman and E. the ground state of the simple harmonic oscillator has an energy 1 ω above the classical ground state energy. there is another source for correction. due to the presence of a Yukawa coupling. We will label the vacuum states |+ . this does not work. |− . Consider the case W (x) = x3 .71) when x0 is a minimum of V (x) so that Ecl = 1 V (x0 ) we have a perturbed 2 energy given by: Epert = Ecl + 1 V (x0 ). this can be made into a very powerful tool. 2 (2.75) This potential clearly has a vacuum degeneracy. . More 2 generally for the Bosonic Hamiltonian: HB = 1 2 (p + V (x)). As classical results are easier to obtain. and Ecl = 0. This is analogous to how the zero point energies of a free oscillator cancel between the Bosons and Fermions. In order to see how nonperturbative eﬀects become relevant we will actually consider the following: W (x) = 1 3 x + ax 3 a < 0. For example.S. Vcl = 1 (3x2 )2 2 G.D. The result is true to all orders in perturbation theory. This generalizes under some circumstances to supersymmetric ﬁeld theories and is known by the name “nonrenormalisation theorem” [26].

80) 0 0 0 1 0 0 0 0 0 0 ψ2 = 0 0 0 −1 0 0 0 −1 · 0 0 0 0 0 0 (2.3 A two variable realization and ﬂat potentials Consider a two variable realization of supersymmetry. The supercharge is given by: 2 Q= α=1 + ψα −pα + i ∂W ∂xα (2.76) (2. This is a tunneling phenomenon (essentially a nonperturbative eﬀect) that is associated with the presence of instantons. 2. its energy gap is: ∆E = a exp − c λ (2.79) where W (x. c are numerical coeﬃcients. y) is a general function of the Bosonic variables x. Note that for a > 0 there are no zero energy solutions already classically.154 Unity from Duality: Gravity.77) E(|A ) − E(|S ) > 0 and the ground state is the symmetric state. Gauge Theory and Strings The energy eigenstates are: |S |A = = 1 √ (|+ + |− ) 2 √ (|+ − |− ) 2 (2. One can realize ψ1 and ψ2 by the following matrices: 0 1 0 0 0 0 0 0 ψ1 = (2.81) The Hamiltonian has the following structure: 0 0 0 H0 0 H11 H12 0 H= 0 H21 H22 0 · 0 0 0 H2 (2. y.78) a.82) .

89) (2. 0) = With.) For n = 0.83) (2. 2 sectors can be solved exactly for E = 0 just as before. 2 For c > 0. 1) = (0. 2 Note. Flat directions in purely Bosonic models are lifted by quantum ﬂuctuations. Classically the ground state is non zero which implies that the supersymmetry is broken. 1 [−∆ + (∇W )2 − ∆W ] 2 1 [−∆ + (∇W )2 + ∆W ] 2 a+1 a+2 |00 F F a+1 |00 . neither ground state is normalisable. Vcl = 1 c2 at y = 0. Berman and E.S.0 ∂1 e−W φ0. What about quantum mechanical eﬀects. can supersymmetry actually be restored? (There are cases where symmetries have been known to be restored quantum mechanically.0 · (2. Consider the potential W = x(y 2 + c) then Vcl = 1 [(y 2 + c)2 2 + 4x2 y 2 ]. Rabinovici: Supersymmetric Gauge Theories The possible states are: (1. We will now show that . Flat directions imply the presence of many vacua.D. 155 (2. In supersymmetric systems the Bosonic and Fermionic ﬂuctuations cancel and the ﬂat directions remain. the potential is 2 the same for all values of x.1 = ∂2 eW φ0.90) Unfortunately.85) H0 H2 = = (2. the answer is no.87) H1 = 1 2 (−∆ + (∇W )2 − ∆W )1 + 2 ∂11 W ∂21 W ∂12 W ∂22 W · (2. What about n = 1 sector? To ﬁnd these states we need to solve the following pseudo-analytic equations: ∂1 eW φ1. (1.86) (2.84) (2. a+2 |00 F F |00 . This is a “ﬂat direction”. Vcl = 1 c2 > 0 which leads to classical SUSY breaking [27].1 = −∂2 e−W φ1. 2 sectors.88) For any W the n = 0. these cannot be solved in general. 1). 0) = (0. Such ﬂat directions appear in abundance in supersymmetric models and lead in some contexts to supersymmetry breaking.

E W (λ) ≥ 0. Gauge Theory and Strings from more general arguments that supersymmetry cannot be restored by quantum eﬀects. F is the fermion number of the state. nF =even (E).92) φ|H|φ · φ|φ (2. The “Witten index”. So we may write: 1 E W (λ) = E W (0) + λ2 + λr2 2 (2.) ∂x W = y 2 + x > 0 .94) where r is some ﬁnite real quantity. ∆ = Tr(−1)F = E nF =even (E) − nF =odd (E) .91) where φ denotes the unperturbed solution. This argument is rather general and also works for the 3 variable potential W = ayz + bx(y 2 + c) (2. thus.156 Unity from Duality: Gravity. (There is no λ dependence from the Yukawa term. [25] is deﬁned by.97) The trace indicates a sum over all states in the Hilbert space. (2. 1 E W (λ) = E W (0) + λ2 + λ 2 d2 x(n · ∇W )φ2 (2. φ2 ≥ 0 and d2 xφ2 = 1.93) (2. nF =odd (E) denote the number . W λ = W 0 + λxi ni where ni ni = 1 and W 0 = x(y 2 + c) is the unperturbed potential. d2 x(n · ∇W )φ2 > 0. EW = Consider perturbing the potential.96) (this is the potential used to break supersymmetry spontaneously in scalar ﬁeld theories.) There is a short more elegant argument for non-restoration of supersymmetry that is based on an index theorem.95) (2. if we now assume E W (0) = 0 then we obtain a contradiction because by taking λ to be small and negative we obtain E W (λ) < 0.

99) (2. Since the Bosons and fermions are paired at all energies greater than zero then ∆ = nF =even (E = 0) − nF =odd (E = 0). Perturbing the spectrum. (2. We have calculated the classical Witten Index to be: ∆cl = 0. Since the index is invariant under perturbations. ∆! 1 = E (2. 2.103) . if ∆ = 0 for some W then there will be no spontaneous symmetry breaking under any perturbations in W . In particular. Returning to the case described above.S. Thus for ∆ = 0 one needs more information.98) If ∆ is calculated for some potential W it will not change under perturbations of W (only En will change).102) We have already shown that nq F =0 (E = 0) = nq F =2 (E = 0) = 0 N N (2.100) w w+! w δ ∆! !=0 w w+! w δ Q Q Q E=0 Q Q Q Q NF= 0 NF=1 NF=0 NF=1 NF=0 NF=1 NF=0 NF=1 Fig. Rabinovici: Supersymmetric Gauge Theories 157 of solutions with Fermi number even/odd with energy E. (2. one obtains: 0 = ∆cl = ∆quantum = nq F =0 (E = 0) + nq F =2 (E = 0) − nq F =1 (E = 0). Berman and E.101) One considers turning on quantum corrections as a perturbation. On the other hand if ∆ = 0 then we do not know whether SUSY may be broken or not since either nF =even = nF =odd = 0 or nF =even = nF =odd = 0.D. N N N (2. ∆ is unchanged.

1. . 1. (2.. ∗ ∗ ∗ {ψi . .. D) = D! · (D − p)!p! (2.111) . 1. ψj } = 0 .109) = = Generically the Dimension of a state with p-Fermions is given by: Dim(p. S= dtgij (x) dxi dxi 1 ¯ + iψγ 0 Dt ψi + Rijkl ψ i ψ k ψ j ψ l . . dt dt 4 (2. {ψi . ψj } = gij . . 2.. .. 1. .4 Geometric meaning of the Witten index Consider a supersymmetry sigma model with a D dimensional target space whose metric is gij [25].. (2. 0 } Dim {|0. ψj } = {ψi ..104) Dt denotes a covariant derivative. Q acts by adding a Fermion hence is a map Q: φp → φp+1 .. 1 Dim = = = 1 D 1 2 D(D − 1) 1. Rijkl is the curvature... (2. there is no supersymmetry restoration.. thus. Gauge Theory and Strings which in turn implies nNF =1 (E = 0) = 0 . (2.108) The Hilbert space may be graded according to the Fermion occupation number: φ0 φ1 φ2 · · φD = = = |0.. i Q∗ = −i i = φ∗ pi i (2..158 Unity from Duality: Gravity.105) (2.110) This is identical to the dimension of p-forms in D dimensions.106) D D Q=i i = φ∗ pi . . 0 } Dim · · |1.107) and pi = −Dxi . 0 Dim {|0...

5 Landau levels and SUSY QM Consider an electron moving in two dimensions in the presence of a perpendicular magnetic ﬁeld. hence the E = 0 states are in the cohomology of Q. The Witten index is given a very physical realization in the following example. (2. H] = 0 .112) The dimension of the cohomology is denoted by the Betti number bp = dimH p . d and the space of states with Fermion occupation number p and supercharge Q.114) The Euler characteristic is a topological invariant and is independent of geometrical perturbations of the manifold. 2. Rabinovici: Supersymmetric Gauge Theories 159 From simple Fermi statistics we see that.118) . This explains why the Witten index is stable against non-singular perturbations. The supercharges are: Q1 Q2 = = 1 √ ((px + Ax )σ1 + (py + Ay )σ2 ) 2 1 √ ((px + Ax )σ2 − (py + Ay )σ1 ) 2 [Qi . (2. Tr(−1)F = χ.116) (2. This is a less familiar realization of supersymmetry as the number of Bosonic and Fermionic oscillators diﬀer. Hence. we have an isomorphism between p-forms with an exterior derivative. It is particular to the quantum mechanical system. Q is nilpotent.113) Recall that to ﬁnd the E = 0 states.117) (2. The Euler characteristic is then given by the alternating sum of the Betti numbers. (2. one solved the equation Qφp = 0. {Qi . 2 (2. H= 1 2 2 (px + Ax ) 1 + (py + Ay ) 1 + Bz σ3 . Thus one has the remarkable correspondence. Berman and E.S. σ3 } = 0.D. y and one Fermion.115) Two Bosons x. As with any nilpotent operator one can examine its cohomology deﬁned by: Hp = {dφp = 0} · {φp = dφp−1 } (2. χ= p (−1)p bp .

In such a case. Take a constant negative B ﬁeld with a = − 4 r2 B. Depend2 ing on the sign of B. y) = exp(−a)f (x + iy) . In a constant magnetic ﬁeld. If B > 0. either φ1 or φ−1 may be normalizable.119) where f and g are arbitrary functions.e. i. .125) The total magnetic ﬂux is a global quantity that is independent of local ﬂuctuations. Multiplying by σ1 gives: ((px + Ax )1 + iσ3 (py + Ay )) One can solve these equations as follows: φ1 (x. To ﬁnd E = 0 states we again use Qφ = 0. n(E=0) = |Φ|.120) φ1 φ−1 = 0 . all n.121) There are an inﬁnite number of E = 0 states. a = − 1 y 2 B. The normalisability of these solutions depends on the function a. (2. 2 From translational invariance. taking for example.) Let us examine this from a topological point of view. Gauge Theory and Strings In Lorentz gauge. The solution is: φ+1 = c n r exp(inθ). Thus the number of E = 0 states is given by the total ﬂux number. (2. (Φ is quantized). take φ1 = 0 and 1 φ−1 = exp − y 2 B g(x − iy). a convenient choice is g(x − iy) = exp(ik(x − iy)). Q is just like the Dirac operator. normalisability requires that n is bounded by [Φ − 1]. The total magnetic ﬂux is: Φ= 1 2π d2 rBz = − 1 2π 2π 0 dθr∂r a |r=∞ . Now assume that Φ is ﬁnite because B is conﬁned in a solenoid. (This is true also for any E not just E = 0. Take Q = Q1 . ∂i Ai = 0 we can write Ai = ij ∂j a Bz = −∇2 a.124) For B constant Φ → ∞ and there are an inﬁnite number of allowed states. Again we have shown how the number of zero energy states does not depend on the local details but only upon global information.122) (2.123) Φ = 0 implies a(r = ∞) = 0 and one has only plane wave normalizable 1 states. (2. for example.160 Unity from Duality: Gravity. 0 (2. φ−1 = exp(a)g(x − iy) (2. r|Φ| (2.

K = x2 4 2 and H is as given above. (2. S= dtL. Berman and E.135) . D] = iK .127) The meaning of D and K is perhaps clearer in the Lagrangian formalism: L= 1 2 g x − 2 ˙ 2 x . t =t+δ (2. t = α2 t (2.130) Symmetries of the action S. K] = 2iD. [H. D] = iH . are given by: t = at + b 1 .133) α 0 0 1 α .132) AT = D acts as dilation AD = 1 0 δ 1 .6 Conformal quantum mechanics 161 Relevant material for this section may be found in [28–30].D. and not the Lagrangian L alone.134) K acts as a special conformal transformation AK = 1 0 δ 1 .1) algebra where: 1 1 D = − (xp + px) . Rabinovici: Supersymmetric Gauge Theories 2.126) is special since g has a meaning.131) A= H acts as translation a c b d . The Casimir is given by: 1 3 g (HK + KH) − D2 = − · 2 4 16 (2.128) (2. [K. This forms an SO(2. t = t · δt + 1 (2. Recall the Hamiltonian: H= 1 2 (p + gx−2 ) 2 (2.129) (2. (2. H is part of an algebra: [H. x (t ) = x(t) ct + d ct + d detA = ad − bc = 1 (2.S.

How do we deal with a system in the absence of a ground state? .137) (2. α− < 0.136) 4 We will now attempt to ﬁnd the zero energy state. 3. The absence of a normalisable ground state for this potential.162 Unity from Duality: Gravity. the spectrum is therefore continuous and bounded from below.139) α=− ± 2 2 This gives two independent solutions and by completeness all the solutions.126 is the open set (0. Take the ansatz (2. The wave functions are given by: √ √ ψE (x) = xJ√g+ 1 ( 2Ex) E = 0. α+ > 0. Thus. there is no normalizable E = 0 solution (not even plane wave normalizable)! H ? ∞ x Fig.138) √ 1 + 4g 1 · (2. (2. φ(x) = xα : H= This implies g = −α(α − 1) solving this equation gives − d2 g + 2 2 dx x xα = 0. Gauge Theory and Strings The spectrum of the Hamiltonian 2. does not lead to a normalizable solution since the function diverges at inﬁnity. Most of the analysis in ﬁeld theory proceeds by identifying a ground state and the ﬂuctuations around it. ∞). is not normalizable either since the function diverges at the origin (a result of the scale symmetry).

Rabinovici: Supersymmetric Gauge Theories 163 One possibility is to accept this as a fact of life. (2. This means we must scale (m − q) → 0. This is associated with the coordinate singularity at the Horizon. At = r · M (2.142) One must interpret what this means physically. Surprisingly this question arises in the context of black hole physics.D.140) This operator has a ground state if v − 4uw < 0. Any choice explicitly breaks Lorentz and scale invariance. Another possibility is to deﬁne a new evolution operator that does have a ground state.144) This produces an AdS2 × S 2 geometry.e. Berman and E. Perhaps it is possible to view this as similar to cosmological models that also lack a ground state such those with Quintessence. The change in evolution operator is now associated with a change of time coordinate. r0 = 1 2 1+ g+ 1 4 · (2. We also wish to keep ﬁxed M 2 (m − q) as we scale M . ds2 = − r M 2 dt2 + M r 2 dr2 + M 2 dΩ2 . 2 (2. The Hamiltonian for this in falling particle in this limit is given by our old friend: g 4l(l + 1) p2 · (2.141) a has dimension of length. G= 1 2 1 K + aH a ≡R (2.S. That is the particle itself becomes BPS in the limit.145) g = 8M 2 (m − q) + H= r + 2 . The black hole is BPS meaning that its mass. The eigenvalues of R are rn = r0 + n . The blackhole metric is given by: ds2 = − 1 + M r −2 dt2 + 1 + M r 2 (dr2 + r2 dΩ2 ) . which we will take by M → ∞ keeping r ﬁxed. . we have g > 0 and there is no ground state. One for which the world line of a static particle passes through the black hole horizon instead of remaining in the exterior of the space time. M and charge. r << M . Consider a particle of mass m and charge q falling into a charged black hole.143) Now consider the near Horizon limit i. Q are related by M = Q. Take for example. 2m 2r M For l = 0. G = uH + vD + wK.

2 (2. S} = g − B − 2iD.148) One now has a larger algebra. the superconformal algebra.147) Representing ψ by 1 σ− and ψ ∗ by 1 σ+ gives the supercharges: 2 2 Q = ψ + −ip + dW dx . Q+ = ψ ip + dw dx · (2. Gauge Theory and Strings 2. S + } = g − B + 2iD = 2K . {Q. {Q .153) and we may solve for the full spectrum: √ 1 ψE (x) = x 2 J√ν (x 2E) .7 Superconformal quantum mechanics The bosonic conformal mechanical system had no ground state.146) 1 − σ3 d2 W dx2 · (2. Again 4 4 the spectrum is continuous though there is no normalizable zero energy . E=0. S = ψ + x . W (x) = yielding a Hamiltonian: H= 1 2 p2 + dw dx 2 1 g log x2 . {Q. This breaking has a diﬀerent ﬂavor from that which was discussed earlier.152) 0 p + 2 g(g−1) x2 0 (2. (2.150) A realization is: B = σ3 .149) (2. neither solution is normalizable. Q+ } {S. H factorizes: 2H = p2 + g(g+1) x2 (2.154) where ν = g(g − 1) + 1 for NF = 0 and ν = g(g + 1) + 1 for NF = 1. The zero energy solutions are exp(±W (x)) = x±g . For this the superpotential is chosen to be.164 Unity from Duality: Gravity. Next the supersymmetric version of conformal quantum mechanics is examined to see if supersymmetry is indeed broken. + (2. S } + = 2H . The absence of a E = 0 ground state in the supersymmetric context leads to the breaking of supersymmetry.151) (2. S + = ψx.

Qα ] = [Pµ .160) (2. Qβ } = {Qα .D.1) (3.32].2) ¯˙ [Pµ . N + } = L+ 4 1 − (H − K ∓ 2iD) 2 T2 |0 = 0 . Berman and E. gB = gsusy (gsusy − 1). its N = 1 version is: µ ¯˙ ¯˙ ¯˙ {Qα . “Ground states” are given by: T1 |0 = 0 . {Qα . H|0 = 0 . 3 Review of supersymmetric models 3. N } = 4 L± = 1 1 R + B − g ≡ T1 2 2 1 1 R + B + g ≡ T2 2 2 1 L− {M + . One needs to deﬁne a whole new set of operators: M =Q−S N =Q +S which produces the algebra: 1 {M. (2. T2 .159) (2. ˙ (3. H have a doublet spectra. One can again deﬁne a new operator with a normalizable ground state. Again we must interpret the absence of a normalizable ground state. This is the supersymmetric analogue of the situation already discussed. Qα } = 2σαβ Pµ . Rabinovici: Supersymmetric Gauge Theories 165 state.S. Qβ } = 0. By inspection the operator (2. M + } = 4 1 {N. (2.158) =Q+S + T1 . Qα ] = [Pµ . The possible particle content of supersymmetric (SUSY) theories is determined by the SUSY algebra.163) A physical context arises when one considers a supersymmetric particle falling into a black hole [31.156) Thus the spectrum diﬀers between the NF = 1 and NF = 0 sectors and so supersymmetry would be broken. .162) + + M + = Q+ − S + N + (2.155) (2.141) can be used provided one makes the following identiﬁcations: NF NF = 1 = 0 gB = gsusy (gsusy + 1) .157) (2. Pν ] = 0.161) (2. N + } = 4 1 {M.1 Kinematics For a more detailed discussion of the material presented in this section see [3]. (2.

they commute with all other charges (they are antisymmetric in ij). ψ) + (λα .4) In the above table. The massive vector multiplet becomes: −1 1 −2 1 −2 0 0 1 2 11 2 (h.8) . 1 2 2 (λα . (3. there is a massless vector multiplet: −1 1 −2 1 −2 0 0 1 2 11 2 (φ. in N = 1. ψ2 ) − 1 . 2. 0. there is again the chiral multiplet which is the same as the massless multiplet but with now massive ﬁelds. and those with dotted indices. 4). Qj α } = 2δ ij σαβ Pµ + δαβ Uij + (γ5 )αβ Vij . In four dimensions we have two component Weyl Fermions. 4) (3.5) λ is a Weyl Fermion and Aµ is a vector ﬁeld. 2). 3 the space-time indices. ψα . 1 ) representation of the Lorentz group. 2 2 (φ. 4). ﬁrst are written the helicities. α 2 ˙ transform under the ( 1 . λα .3) U and V are the central charges i.7) 0 2 For massive multiplets. φ denotes a complex scalar and ψ a Weyl Fermion. − .e. Aµ ) (4. (3. 1 2 (4. Aµ ) (4. We will discuss in this section the d = 4 realisations with.166 Unity from Duality: Gravity. 0 . They are associated with BPS states such as monopoles. The simplest is the the chiral multiplet: 1 1 − . Gauge Theory and Strings This algebra can be generalized to include a higher number of supersymmetries N = 2. 0. and ﬁnally are the number of physical degrees of freedom carried by the Bosons and Fermions. ψ) (2. or β 2 Consider ﬁrst the massless representations of N = 1 supersymmetry. . 4 by: µ ¯ {Qi . ν = 0.6) and a massless hypermultiplet which is given by: 0 1 − 2 0 1 2 (φ1 . With N = 2 supersymmetry. ψ1 ) + (φ2 . (3. 1. 0) representation. Aµ ) (2. The vector multiplet has: 1 1 −1. Those with α or β indices transform under the ˙ (0. µ. 2) (3. ˙ ˙ ˙ α ˙ (3. then the associated component ﬁelds.

1 .10) The following identity will be of use: ¯ d2 θd2 θL = ∂2L d2 θ ¯ ¯ · ∂ θ1 θ2 (3. Recall the integration formulas for Grassmann variables: dθα θα = ∂ =1.. (3. −2 0 2 1 −2 0 1 2 0 where I = 1. ˙ ∂θα ∂ µ ¯˙ Qα = − ¯ + iθα σαα ∂µ . massive vector multiplets may appear by a supersymmetric analogue of the Higgs mechanism. the massless vector multiplet is: 0 1 −2 0 1 2 1 1 −2 0 2 −1 . ˙ ∂ θα ˙ (3. 8) (3.D. Rabinovici: Supersymmetric Gauge Theories 167 Where h is a real scalar ﬁeld. This can occur dynamically.4.14) .12) It will also be useful to deﬁne a supercovariant derivative: Dα = ∂ µ ¯˙ + iσαα θα ∂µ .6.. (3. . ∂θα dθα = 0. ˙ ∂ θα ˙ ¯˙ Dα Φ = 0.S. θα .11) The supercharges can be realized in superspace by generators of supertranslations: Qα = ∂ µ ¯˙ − iσαα θα ∂µ . Constructing a Lagrangian out of superﬁelds provides a useful way to construct explicitly supersymmetric Lagrangians. φI . Aµ ) (8. Berman and E. 3. θα . The massive vector multiplet has a diﬀerent ﬁeld content than the massless vector multiplet because a massive vector ﬁeld has an additional physical degree of freedom. 1 (λa . ˙ ∂θα ∂ µ ¯˙ Dα = − ¯ − iθα σαα ∂µ . a = 1.2 Superspace and chiral ﬁelds ¯˙ Spacetime can be extended to include Grassmann spinor coordinates. One sees that the massive vector multiplet is composed out of a massless chiral plus massless vector multiplet. Superﬁelds are then functions of this superspace. With N = 4 supersymmetry.9) 1 .13) A superﬁeld Φ is called “chiral” if: (3.

19) is a invariant under supersymmetric transformations (up to a total derivative). ¯¯ (3. The product of two chiral ﬁelds also produces a chiral ﬁeld.15) (3.16) The Taylor expansion terminates because of the anticommuting property of the Grassmann coordinates. σ 2 (3. ¯ y µ = xµ + iθσ µ θ which produces the following expansion for a chiral ﬁeld. These are called the component ﬁelds of the superﬁeld. √ Φ(y) = A(y) + 2θψ(y) + θθF (y). (3.23) . any polynomial. ψ(x) will be associated with a Weyl Fermion and F (x) acts as an auxiliary ﬁeld that contributes no physical degrees of freedom. After the integration some terms will disappear from the expansion of Φ(x) leaving only: √ (3. (3. W (Φ) can be used to construct a supersymmetry invariant as L= d2 θW (Φ) = FW (Φ) (3.22) ¯¯ After expanding and extracting the θθθθ term is (up to total derivatives): i ¯ Fi∗ Ff − |∂µ A|2 + ∂µ ψ¯ µ ψ. The kinetic terms are described by: ¯¯ ¯ d2 θd2 θΦi Φj = Φi Φj |θθθθ .20) Φ(x) = A(x) + 2θψ(x) + θθF (x) A(x) will be associate with a complex Boson. This is used to provide a potential for the chiral ﬁeld. Therefore.17) 4 i µ¯ − √ θθ∂µ ψ(x)σ θ + θθF (x).18) 2 The key point is that L= d2 θΦ(x) (3.21) is a supersymmetry invariant.168 Unity from Duality: Gravity. As a function of x it may be written as follows: Φ(x) = √ 1 ¯¯ ¯ A(x) + iθσ µ θ∂µ A(x) + θθθ θ A(x) + 2θψ(x) (3. Gauge Theory and Strings One introduces the variable.

At the ground state this must vanish i. The equation of motion ∗ for Fk is as follows: ∗ Fk + λ∗ + mij A∗ + gijk A∗ A∗ . VF (A∗ .25) One must now eliminate the auxiliary ﬁelds Fi .28) ¯ K(Φ. (Note. For the Φ ¯ usual kinetic terms. A) = 0. VF (A∗ . A). .e. in fact it is true to all orders in perturbation theory as there exists a non renormalization theorem for the eﬀective potential. Berman and E. Rabinovici: Supersymmetric Gauge Theories One thus composes the following Lagrangian: L = = 1 1 ¯ Φi Φi |θθθθ + λi Φi + mij Φi Φj + gijk Φi Φj Φk ¯¯ 2 3 169 (3. For the case of a sigma model with a target space whose metric is gij . VF ≥ 0). Φ) is called the K¨hler potential. 3. this metric is related to the K¨hler a potential by: gij = ∂2K · ¯ ∂ Φi ∂Φj (3.29) The above supersymmetry invariant (3.28) which previously gave the usual kinetic terms in the action.24) θθ 1 ¯¯ i∂ ψi σ ψi + A∗ Ai + Fi∗ Fi + λi Fi + mij Ai Fj − ψi ψj i 2 +gijk (Ai Aj Fk − ψi ψj Ak ) + h.3 K¨hler potentials a To describe the supersymmetric Lagrangian for scalar ﬁelds spanning a more complicated manifold it is convenient to introduce the following supersymmetry invariant: ¯ d4 θK(Φ.S. Φ).27) where the last term is a potential for A. k i i j (3. This in turn implies that Fi = 0 for the ground state. (3. with the target space metric given by equation (3. (3. One may add any function of Φ or a ¯ to the integrand since these terms will vanish after integration.26) This gives: L 1 1 ∗ ∗ ¯¯ = i∂ ψi σ ψi + A∗ Ai − mij ψi ψj − m∗ ψi ψj i 2 2 ij ∗ ¯ ¯ −gijk ψi ψj Ak − gijk ψi ψj A∗ − Fi∗ Fi k ∗ (3.c.29).D. produces for general K the action of a supersymmetric sigma model. A . Although this is a classical analysis so far. these are known as the F terms. Fi∗ . K is taken to be given by K = ΦΦ which produces the ∗i µ j −δij ∂µ A ∂ A kinetic terms for the scalars.

x).5 Global symmetries R-symmetry is a global U (1) symmetry that does not commute with the supersymmetry. Φ∗ . x) ¯ exp(−2inα)Φ(exp(iαθ). RΦ(θ. (3. 3.32) and hence there will be a choice of vacuum: either A = or A = 0. 2 3 To show that the form of W remains invariant. the following ingredients are used: the holomorphic dependence of W (Φ. the R-charges of the component ﬁelds may be read oﬀ as follows: R(lowest component of Φ) = R(A) ≡ n . λ). where F1 = −λ − mA + gA2 . Note. λ∗ .35) Since the R-charge does not commute with the supersymmetry. The claim is that the form of the eﬀective interacting superpotential Weﬀ will be the same as the classical superpotential Wcl . R(ψ) = n − 1 . This implies the classical result is exact.4 F-terms In this section we examine the vanishing of the potential generated by the F terms. m. Take Wcl to have the form: 1 1 (3. (3. The Witten index Tr(−1)F = 2. R(F ) = n − 2. (3.170 Unity from Duality: Gravity. x) = exp(2inα)Φ(exp(−iαθ). (3. The action of the R-symmetry on a superﬁeld Φ with R-character n as follows. There is no supersymmetry breaking classically. For a superﬁeld Φ with R-character n. and are there other internal symmetries broken. Take the example of the one component WZ model.34) (3. and the global symmetries present in the theory [33]. is there a non renormalization theorem.36) .31) This has a solution. VF = 0 ⇔ Fi = 0 ∀i . Gauge Theory and Strings 3. V = A∗ A|(gA − m)|2 m g (3. the component ﬁelds of the chiral ﬁeld have diﬀerent R-charges. they have a solution. x) = ¯ ¯ RΦ(θ. that is W is independent of m∗ . that for λ = 0.33) W = mΦ2 + λΦ3 . Generically. The issues we are interested in are whether supersymmetry is spontaneously broken.30) are n (complex) equations with n (complex) unknowns.

(3.40).S.38) to have zero R charge requires that R(W ) = 2. R(dθα ) = −1 171 (3. All component ﬁelds are charged the same with respect to this U (1) symmetry.42) λ 3 Φ 3 (3. There is also one other global U (1) symmetry that commutes with the supersymmetry. The expectation value inherits the quantum numbers of the ﬁeld. barred variables having opposite R charge.43) These symmetries are next used to prove the nonrenormalisation theorem. The charges are summarized in the following table: Φ m λ W U (1) U (1)R 1 1 −2 0 −3 −1 0 2.37) ¯ with. This can be viewed as simply a book keeping device or more physically one can view the coupling as the vacuum expectation value of some ﬁeld. Adding the cubic term: W3 = produces V = |λ|2 |A|4 + λAψψ. Demanding that the terms in the action maintain this symmetry requires an assignment of U (1) charges to λ. The kinetic term ΦΦ is ¯θθθ is an invariant. . R(ψ) = 0.40) (3. (θ d2 θW (3. To restore R-invariance requires λ is assigned an R-charge of −1.D.39) This term is not R-invariant with the R-charges given by (3. Rabinovici: Supersymmetric Gauge Theories The R-charge of the Grassmann variables is given by: R(θα ) = 1 . Berman and E. For the mass term W = 1 2 2 mΦ . to have vanishing R-charge requires R(Φ) = R(A) = 1 . W = mψψ + m2 |A|2 .41) (3. This is how one treats for example the mass parameters of Fermions in the standard model.) For the potential term. ¯ an R invariant. and m. (3.

36] is the ﬁeld theory analogue of the potential given by equation (2.96) for supersymmetric quantum mechanics. (On the other hand. (3. mΦ2 g λΦ m = λn Φn+2 1 1 mΦ2 + Φ3 + an · 2 3 mn−1 n=2 ∞ λΦ m · (3. The following anecdote may be of some pedagogical value [34]. (3. The presence in the collaboration of a student who paid little respect to the general counter arguements made the discovery possible. λ) = g1 (mΦ2 . This requires that n ≥ 0.45) (3. Gauge Theory and Strings 3. The Fayet-O’Raifeartaigh potential [35. Supersymmetry resisted breaking attmepts for both theories of scalars and gauge theories.172 Unity from Duality: Gravity.46) As λ → 0 one must recover the classical potential.c. if he classical superpotential has unbroken supersymmetry so does the full theory. In particular. m.7 Supersymmetry breaking Let us examine now how supersymmetry may be spontaneously broken.44) combining this with the invariance under the U (1) R-symmetry implies Weﬀ = mΦ2 g Consider expanding in |λ| << 1. There can therefore be no corrections to the form of Weﬀ [33].47) . It turns out that a short time after supersymmetry was introduced arguments were published which claimed to prove that supersymmetry cannot be broken spontaneouly. One could be surprised that the breaking was ﬁrst achieved in the gauge systems. the Kahler potential is renormalized. Consider the limit m → 0. In order to break supersymmetry a minimum of three chiral ﬁelds are needed: LPotential = λΦ0 + mΦ1 Φ2 + gΦ0 Φ1 Φ1 + h. This was done by Fayet and Illiopoulos. λΦ3 ) .) 3. Fayet went on to discover the breaking mechanism also in supersymmetric scalar theories as did O’Raighfeartaigh.6 The eﬀective potential That the eﬀective potential be invariant under the global U (1) implies that Weﬀ (Φ. for the function to be free of singularities implies that n < 1.

D. We turn next to supersymmetric theories that are gauge invariant. 0.48) (3. Wα .53) B. χ. The symmetry is actually U (1)C as opposed to the usual U (1)R because although the vector ﬁeld transforms with a real gauge parameter.49) (3.51) (3. and the Fermions have masses: 0. Doing so one discovers that in the ground state A1 = A2 = 0 and A0 is arbitrary. like in the quantum mechanical example (2. 4 (3. and C are gauge artifacts and can be gauged away.52) (3. Λ an arbitrary chiral ﬁeld: ¯ V → V + i(Λ − Λ) (3. V = 1 2 component ﬁelds. Aµ are real and C is complex.55) . It obeys a ¯ ¯ ¯ ¯¯ B + θχθ χ + +θ2 C + θ2 C − θσ µ θAµ 1 ¯ ¯ ¯ 1¯ +iθ2 θ(λ + σ µ ∂µ χ) − iθ2 θ(λ − σ µ ∂µ χ) ¯ 2 2 1 ¯ + θ2 θ2 (D2 + ∂ 2 B) 2 (3. There will be a local U (1) symmetry with gauge parameter. To ﬁnd the ground state one must write out the full Lagrangian including the kinetic terms in component ﬁelds and then minimize.96). Berman and E. Rabinovici: Supersymmetric Gauge Theories Minimizing the potential leads to the following equations: 0 = 0 = 0 = λ + gΦ1 Φ1 mΦ2 + 2gΦ0 Φ1 mΦ1 .8 Supersymmetric gauge theories A vector superﬁeld contains spin 1 and spin ¯ reality condition V = V . The zero mass Fermion is the Goldstino. Computing the masses by examining the quadratic terms for component ﬁelds gives the following spectrum: the six real scalars have masses: 0.S.50) These cannot be consistently solved so there cannot be a zero energy ground state and supersymmetry must be spontaneously broken. the other ﬁelds transform with gauge parameters that depend on the imaginary part of Λ. Dβ Wα = 0. D. 2m. It is possible to construct a chiral superﬁeld.54) B. m2 ± 2gλ. 3. The arbitrariness of A0 is a ﬂat direction in the potential. 173 (3. m2 m2 . from V as follows 1¯¯ ¯˙ Wα = − DDDα V .

61) D is clearly a non propagating ﬁeld.c. λ → λ . Fµν its ﬁeld strength.54).60) Expanding this in component ﬁelds produces. A monopole in the presence of such a coupling will get electric charge through the Witten eﬀect. 4g 2 2g g 32π 2 (3. W is gauge invariant. The supersymmetries acting on the component ﬁelds are. 1 ¯ ¯ ¯¯ ¯ V (y) = −θσ µ θAµ + iθ2 θλ − iθ2 θλ + θ2 θ2 D (3.57) ¯ α 2 Where Aµ is the vector ﬁeld.58) Note.67) One may also add to the action a term linear in the vector ﬁeld V .62) δA = √ µ √ δ ψ = i 2σ ¯∂µ A + 2 F (3. the component ﬁelds transform under a now U (1)R symmetry as: Aµ → Aµ − i∂µ (B − B ∗ ) . The following supersymmetric gauge invariant Lagrangian is then constructed: L= d2 θ −iτ 16π W α Wα + h. Gauge Theory and Strings One may choose a gauge (called the Wess Zumino gauge) in which B.174 Unity from Duality: Gravity. τ= θ 4π +i 2 · 2π g (3.68) .c. D → D.59) where the coupling constant τ is now complex.56) 2 i β ¯ Wα (y) = −iλα + δα D − (σ µ σ ν )β Fµν θβ + (σ µ ∂µ λ)α θ2 . (up to total derivatives): √ 2 ψ (3. λ is the spin 1 ﬁeld and D 2 is an auxiliary scalar ﬁeld. L= 1 i θ −1 ¯ Fµν F µν + 2 D2 − 2 λσDλ + (∗F )µν Fµν . (3. Under the symmetry (3. (3.64) ¯ + ¯σµ ∂ν λ) − (µ ↔ ν) δ Fµν = i( σµ ∂ν λ (3.66) (3. (3. (3.63) √ µ δ F = i 2¯σ ∂µ ψ ¯ (3. ¯ µ µ (3. The theta term couples to the instanton number density (this vanishes for abelian ﬁelds in a non-compact space). C and χ = 0 and then expand in terms of component ﬁelds.65) ¯ δλ δD = = i D + σ µν Fµν ¯ ¯σ ∂µ λ − σ ∂µ λ . known as a Fayet-Iliopoulos term [37]: 2K ¯ d2 θd2 θV = KD = dθα Wα + h.

69) Under the gauge transformation ¯ V → V + i(Λ − Λ) .76) (3. the D and F ﬁelds.c. = 0.75) The ﬁrst term is the F -term and the second is the D-term. Some remarks about this potential: Generically the F -terms should vanish since there are n equations for n unknowns. (3. Berman and E. Consider ﬁelds Φ1 .) + Φ1 exp(eV )Φ1 4 ¯ +Φ2 exp(−eV )Φ2 + mΦ1 Φ2 + h. (3.D.61) for the vector ﬁeld and a potential. Both these terms need to vanish for unbroken supersymmetry. Φi → exp(−iqi Λ)Φi . L= d2 θ W α Wα + ¯¯ ˜ d2 θΦi exp(qi V )Φi + W (Φ) (3. (3. Φ2 with opposite U (1) charges and Lagrangian given by: L = 1 ¯ (W α Wα + h.72) (3. These ideas are demonstrated by the following example.71) Writing out the term (3. ˜ Adding the kinetic term (3. that is if the U (1) is not spontaneously broken then supersymmetry is broken if and only if KF. 2 2 F ∗ F − |∂µ φ + (3. (3.S.I.73) There are two auxiliary ﬁelds.74) this produces the following potential. In order that the theory is free from chiral anomalies one requires: qi = 3 qi = 0.c. W (Φ) for the matter.69) in components produces: L = iq iq ¯σ Aµ φ|2 − iφ¯ ∂µ + qAµ ψ 2 2 iq 1 ¯ ¯¯ ¯ − √ φλψ − φλψ + qDφφ. When K = 0 and the F -terms have a solution then so will the D-term and there will be no supersymmetry breaking. gives the total Lagrangian. (3. V = i | ˜ ∂ W 2 q2 (2K + | + ∂φi 4 |φi |2 2 . Rabinovici: Supersymmetric Gauge Theories 175 Its role will be discussed later. The U (1) gauge ﬁelds couple to charged chiral matter through the following term L= i ¯¯ d2 θd2 θΦi exp(qi V )Φi .70) Since there are chiral Fermions there is the the possibility for chiral anomalies. If φi = 0. + 2KV.77) .

The potential expanded around this minimum ˜ becomes. The Higgs ﬁeld. Gauge Theory and Strings This leads to the potential: V = where e D + K + (A∗ A1 − A∗ A2 ) = 0 2 2 1 F1 + mA2 ∗ = 0 F2 + mA∗ = 0. The vector ﬁeld has zero mass.82) Consider the case. 1 2 ∗ ∗ D + F1 F1 + F2 F2 2 (3. Next.79) (3. For K = 0. √2 (A + A∗ ))2 has the same mass as the photon.81) The scalars have mass. 1 Leading to the following expression for the potential: 1 1 1 1 2 V = K 2 + m2 + eK A∗ A1 + m2 − eK A∗ A2 + e2 (A∗ A1 −A∗ A2 ) . the Goldstino. 2 The ﬁrst term implies that supersymmetry is broken for m > 0. m2 < 1 eK.78) (3. consider the case. m2 > 1 2 eK. A2 = 2 v where v 2 ≡ 4 2 e2 . Note that a trace of the supersymmetry remains as TrMB = 2 TrMF even after the breaking of supersymmetry. 1 2 1 2 2 2 2 8 (3. m2 + 1 eK and 2 m2 − 1 eK. The photon is now massive. with A ≡ A1 and A ≡ A2 − v: V = 2m2 1 (eK − m2 ) + e2 2 +2m2 A∗ A = + 1 2 2 e v Aµ Aµ 2 1 ˜ ˜ √ A + A∗ 2 2 1 eK−m2 (3.85) 1 1 2 2 ( e v ) 2 2 1 ˜ ¯¯ ˜ ¯ m2 + e2 v 2 ψ ψ + ψ ψ + 0 × λλ.83) (3.176 Unity from Duality: Gravity.84) (3.) The massless Fermion (the Photino) is now a 2 Goldstino. m2 = 1 e2 v 2 implying that the U (1) symmetry is broken γ 2 1 ˜ ˜ as well. supersymmetry is broken as the Bosons and Fermions have diﬀerent masses. at the minimum. Since the vector ﬁeld remains massless then the U (1) symmetry remains unbroken. (For K = 0 though the symmetry is restored. Two Fermions have non-zero mass and there is one massless Fermion.80) (3. A1 = 0. . Two Fermions have mass m 2 and one Fermion is massless.

λa . This gives a U (1) theory with oppositely charged ﬁelds Φ+ . a = − which leads to no U (1) breaking and a+ a− = − 1 8 λ− mµ µ . a− → e−αa− (3. the presence of a ﬂat direction: a+ → eα a+ .88) λ m (3. The vanishing of the D-term implies: D = |Φ+ |2 − |Φ− |2 = 0. Berman and E.87) which breaks the U (1) symmetry. Aa . Four real ﬁelds obeying the constraint: Φ+ = exp(iα)Φ− (3. Only if there is a U (1) factor will the Fayet-Iliopoulos term be non-vanishing. Since there is a broken global symmetry this would lead to the presence of Goldstone Boson corresponding to the broken U (1)R . Typical generic supersymmetry breaking leads to the breaking of R-symmetry. Da . We wish to examine the properties of potentials with ﬂat directions.a = − g g (3.): a+ a− = 0 . a = 1.90) ..S. So far we have only dealt with U (1) vector ﬁelds.91) (3. the index a is the group index. 2 (3. Inverting this argument leads to the conclusion that supersymmetric breaking in nature cannot be generic since we do not observe such a particle. Next consider a more generic example where there is U (1) breaking but no supersymmetry breaking. Φ is neutral under the U (1) while Φ+ has charge +1 and Φ− has charge −1. The ﬁelds are in an adjoint representation of the group. Rabinovici: Supersymmetric Gauge Theories 177 In the above example there is both supersymmetry breaking and U (1) symmetry breaking except when m = 0 in which case the supersymmetry remains unbroken.D. Note.86) There are two branches of solutions to the vacuum equations (a denotes the vacuum expectation value of A etc. Φ− . Take the simple example of an abelian theory with no Fayet-Iliopoulos terms and no F -terms. The potential is given by: L= 1 mΦ2 + µΦ+ Φ− + λΦΦ+ Φ− + h.c.dim(group) and Da = µ ¯ a i φi TR(Φi ) Φi . One can also consider non-abelian groups.89) leaves a− a+ ﬁxed and the vacuum equations are still satisﬁed.

Note. Z2 . . Φ + = Φ− . The metric will then be: ds2 = ¯ 1 dM dM √ · ¯ 2 MM (3. (as is the case above) the moduli space is determined in the following way. This leaves a two dimensional moduli space. then provided there are solutions to the F -terms equations. Φi Moreover the moduli space. The moduli space is the space of all ﬁelds Φi . GC is spanned by a basis of the independent gauge singlets (such as m = φ+ φ− ). The moduli space is then given by the space of ﬁelds that solve the F -term equations modulo complex GC gauge transformations. By expanding the D-term. the D-terms automatically vanish. What is the physical interpretation of this singularity in moduli space? Let us study the space in terms of gauge invariant variables: M ≡ Φ+ Φ− . (3.93) (3. However there is still the U (1) gauge symmetry that one must mod out by. modulo GR gauge transformations. When K = 0 but in the presence of an F -term. A convenient gauge ﬁxing is. There remains a Z2 residual gauge symmetry however: Φ+ → −Φ+ . There is therefore a ﬁxed point at Φ+ = 0.178 Unity from Duality: Gravity. the Kahler potential becomes: ¯ ¯ ¯ ¯ K = Φ+ Φ+ + Φ− Φ− = 2Φ+ Φ+ = +2 M M . Using Φ− = Φ+ . ones sees that M is a parameter that determines the mass of the matter ﬁelds. When K = 0.95) (3.94) There is a singularity in moduli space when M = 0. In the absence of an F -term then there is always a solution to the D-term equations. there is no nonrenormalisation theorem for the Kahler potential. Gauge Theory and Strings implies a three dimensional conﬁguration space.92) After moding out by this action. Thus the singularity in moduli space is a signature of a particle becoming massless. the classical moduli space is given by the C orbifold. which will be a singularity. such that Da = 0. This is equivalent to the space of all constant Φ ﬁelds modulo the complex GC gauge transformations.96) (3.

3) This lead to a tilted lattice of dyonic charges and one may condense dyons with charges (n0 .S. n2 T + C m2 < · T N (4. T = N g · 2π For a given theory there is a lattice of electric and magnetically charged operators where the electric charge is denoted by n and the magnetic charge by m. How does this relate to QCD? There are ideas that conﬁnement in QCD occurs due the condensation of QCD monopoles [45–47]. an electric charge picks up a magnetic charge and becomes dyonic [44]. m) is perturbative i. Magnetic charges are screened. m). electric charges are conﬁned and the log of the Wilson loop is proportional to the area. this is the dual of electric condensation.e. How does the theory change? There are three phases depending which operators condense. F > 0. It is diﬃcult to . 2π (4. n =n+ θ m. In the presence of a theta parameter. This leads to what is called oblique conﬁnement with the charges commensurate with (n0 . magnetic charges are conﬁned. C ﬁxed and vary T . T N (4. (This is called the Higgs phase.) At high T . there is electric condensation which implies that there is electric charge screening. Rabinovici: Supersymmetric Gauge Theories 4 Phases of gauge theories 179 First we will gain our intuition from D = 4 lattice gauge theories. m0 ). magnetic condensation occurs.1) however when the free energy is negative for the operator (n. m0 ) being screened and all other charges being conﬁned [43].2) Keeping N . and the log of the Wilson loop is proportional to the length of the perimeter of the loop. that is. ZN valued gauge ﬁelds with coupling g [38–43] 2 The eﬀective “temperature” of the system will be given by. this is the Coulomb phase. it condenses indicating the presence of a relevant operator and hence an infra-red instability.D. it is an irrelevant operator and weakly coupled to system.e. so long as the free energy. this occur when. i. Berman and E. An operator with charges (n. (This is called the conﬁnement phase. n2 T + C m2 > . At small T .) For intermediate values of T it is possible that there is neither screening of charges nor conﬁnement.

Novel properties of these theories have been discovered such as new types of conformal ﬁeld theories and new sorts of infra-red duality. 2 supersymmetries. This has been achieved for gauge theories with N = 1.180 Unity from Duality: Gravity. The diﬀerent possible phases. The Lagrangian is: L = (−iτ )TrW α Wα d2 θ + h. There are many new methods that have been utilized and the phase structures of these theories have been well investigated [15. To this we turn next. ˜ ˜ +Q+ exp(−2V )QF + Q exp(2V )Q+ |θθθθ + mF QF Q|θθ (5. study this phenomenon directly. adding scalars and making the the theory supersymmetric one can calculate the condensation of monopoles in a four dimensional gauge theory. and NF chiral multiplets in the fundamental representation of SU (NC ) and NF chiral multiplets in the antifundamental representation. 4. however by embedding the monopole in a spontaneously non-abelian theory with an additional scalar ﬁeld one may smooth out the core of the monopole and remove the singularity. One may proceed analogously.c.1) F F ¯¯ . Consider the case of an N = 1 vector multiplet with gauge group SU (NC ). 48. 49]. The Dirac monopole in a U (1) gauge theory is a singular object. Gauge Theory and Strings m (magnetic) Higgs (T<<1) Possibly Coulomb (T~1) (T>>1) Confinement n (electric) Fig. by enriching QCD. 5 Supersymmetric gauge theories/super QCD The goal will be to examine theories that are simple supersymmetric extensions of QCD.

Possible phases of gauge theories (g1 and g2 are some relevant/marginal couplings).D. R(λ) = 1. Qi a ˜ Qa i (5.5) (5. 5. Rabinovici: Supersymmetric Gauge Theories 181 g2 2) SU( GS HIG U(1 ) CO UL OM B SU (3) NT EME NFIN CO g1 Fig. 2NF × RF . while RF is now chosen so that the total anomaly vanishes. This leads to RF = − NC · NF (5. where the coupling is: τ= 4π θ +i 2 · 2π g (5. This is denoted as U (1)R and is a full quantum symmetry. Berman and E. The Chiral Fermions contribute.S. 2NC + 2RF NF = 0. The adjoint Fermion contributes 2NC × R(λ) to the anomaly. SU (NF )L NF 1 1 × SU (NF )R 1 ¯ NF 1 × U (1)V 1 −1 0 × U (1)A 1 1 0 × U (1)RC 1 1 1. because 2 ∼ ¯ the global ﬂavor symmetry is enhanced 2.4) . A single U (1) symmetry survives the anomaly. when NC = 2. There is an anomaly of the U (1)A × U (1)R symmetry. to SO(2NF )L × SO(NF )R .3) Wα Note. the ﬁelds are charged under the following global symmetries.2) Apart from the local SU (NC ) gauge symmetry.

11) For NF = 0 or for NF = 0 and mF = 0. the rank of the group +1. There are 2NC Fermionic zero modes (from the vector multiplet). The gluino condensate is: λλ = exp 2πik NC Λ3 C N (5. Tr(−1)F = NC i. This leads to the following non-anomalous global charges: SU (NF )L Qi a ˜ Qa i Wα NF 1 1 × SU (NF )R 1 ¯ NF 1 × U (1)V 1 −1 0 × U (1)R NF −NC NF NF −NC NF (5. This leaves NC vacua. RB = 1 − NC NF − NC = · NF NF (5. there is no moduli space.12) .9) (5. the vacuum structure is an infra-red property of the system hence having mF = 0 is equivalent to setting NF = 0 in the infra-red. Consider the quantum moduli space of the case where NF = 0.8) 1. Further breaking occurs because the gluino two point function acquires a vacuum expectation value which breaks the symmetry down to Z2 . NF (5.7) where RC is the classical R-charge and A is the classical U (1)A charge. Note. This indicates there is no supersymmetry breaking.6) The non-anomalous R-charge. is given by: R = RC − NC QA .e. One is now ready to identify the classical moduli space. 5. These Fermionic zero modes break through instanton eﬀects the original U (1)R down to Z2NC . The Witten index.10) (5. + (5. Thus.1 The classical moduli space The classical moduli space is given by solving the D-term and F -term equations: Da ¯ FQF ¯˜ FQF ˜ ˜ = Q+ T a QF − QF T a QF F ˜ = −mF Q = −mF Q. Gauge Theory and Strings The Bosons in the chiral multiplet have an R charge one greater than the Fermions in the multiplet.182 Unity from Duality: Gravity.

0 0 a2 0. NC −1 label the vacua. The number of massless vector Bosons becomes 2 2 NC − ((NC − NF )2 − 1) = 2NC NF − NF ... (5. By setting some subset of ai to zero one may break to a subgroup of SU (NC ) that is larger than SU (NC − NF ).15) the number of massless scalar ﬁelds becomes.0 (5. (To validate this classical analysis the vacuum expectation values must be much larger than any dynamically generated scale.. Further details of quantum moduli spaces will be discussed later. This is called the Higgs phase.) The gauge symmetry is partially broken: SU (NC ) → SU (NC − NF ).13) · 0. The classical moduli space is determined by the following solutions to the D-term equations: a1 0 0. .16) The matrix ˜ M˜ ≡ Q˜Qj ij i forms a gauge invariant basis.14) This is for generic values of ai .S. ai . There are NF diagonal non-zero real entries... .0 ˜ Q=Q= · 0.e.e.. Chiral symmetry breaking produces a mass gap. Note.18) (5. Berman and E.17) and as before when singularities appear i. Consider the case where mF = 0 and 0 < NF < NC . 2 2 2NC NF − (2NC NF − NF ) = NF .0 N ×N . i. Rabinovici: Supersymmetric Gauge Theories 183 where ΛNC is the dynamically generated scale of the gauge theory and k = 1. (5.D. Also.. ai >> Λ. because chiral symmetry is discrete there are no Goldstone Bosons. F C Where the row indicates the the ﬂavor and the column indicates the colour. (5. ¯ K = 2Tr (M M ) (5.0 aNF 0. The Kahler potential is then. if NF = NC − 1 then the gauge group is complete broken. detM = 0 this signals the presence of enhanced symmetries.

19) with the constraint that a |ai |2 − |˜i |2 = ρ. Generically. B˜ = i F ˜ i1 i1 ···iNC Qj1 ˜ ˜ iN · · · QjNC .) There is a generic breaking of gauge symmetry. when ˜ ai = ai = 0 then a subgroup of the SU (NC ) can remain. Nf = NC + 1. 0 ˜= Q 0 · 0 a2 ˜ · · 0 · 0 ·· ·· ·· ·· ·· ·· aNC ˜ 0 0 0 NF ×NC (5. We will now consider some special cases. 2(NC + 1) degrees of freedom given by: j Bi = i1 ii1 ···iNC Qj1 iN ˜ · · · QjNC . B and B are classically dependent. This leaves only NF +1 independent moduli. Gauge Theory and Strings When NF ≥ NC . The dimension of the moduli space is given by: 2 2 2 2 2NC − (NC − 1) = NC + 1 = NF + 1. consider the case.20) Generically the SU (NC ) symmetry is completely broken. F (5. again there are NF moduli from Mi˜ . F iN (5. a a 0 ˜ 0 1 1 0 Q= 0 · 0 a2 · · 0 · 0 ·· ·· ·· ·· ·· ·· aNC 0 0 0 NF ×NC . However. 2 NF (5. ﬁrst the classical moduli space for NF = NC .26) 2 giving again an NF + 1 dimension moduli space. (The moduli space is not smooth. as well as the gauge symmetry being completely broken the global ﬂavor symmetry is also broken. There is a singular point in the ˜ moduli space where M = 0 = B = B.23) 2 ˜ which means M .184 Unity from Duality: Gravity.22) There is however a classical constraint: ˜ detM − B B = 0 (5. 2 Next. one has the following classical moduli space. (5.21) There are degrees of freedom from M˜ and naively one would have two ij further degrees of freedom from: B= i1 i1 ···iNC Qj1 iN ˜ · · · QjNC . B = F ˜ i1 i1 ···iNC Qj1 ˜ · · · QjNC .25) (5.24) However there are again the classical constraints: detM M˜ B ji i − = Mi˜ B i B j = 0 j Mi˜ B = 0 j ˜ j (5. There are also. .

Berman and E. the dynamically generated QCD scale is assigned charges as m and g were before. The symmetries imply.S. Q.27) Λ. 3NC < NF < 3NC . The following table summarizes the charges under the various groups [51–53]. the charges of W for the ﬂavor symmetries. NF = NC . These restrictions ﬁx: W =c Λ3NC −NF detM 1 NC −NF · (5.28) a. NF > 3NC . NF = NC + 1. the classical moduli space gets completely lifted by these nonperturbative eﬀects. NC + 1 < NF ≤ 3NC .29) For non vanishing c.3 Quantum moduli space for 0 < NF < NC 2 ˜ Classically. The power 3NC − NF is the coeﬃcient in the one loop beta function. . c is a numerical coeﬃcient. b are to be determined. If one requires the U (1)A charge to vanish then this implies a = −b. the superpotential. If c does not vanish. Automatically. One examines the charges of W under the various symmetries. the dimension of the moduli space is NF from Q.2 Quantum moduli spaces 185 One is required to examine on a case by case basis the role that quantum eﬀects play in determining the exact moduli space.D. SU (NF )L × SU (NF)R and the U (1)V vanish. all the moduli are now lifted and there is no ground state. 5. Quantum eﬀects both perturbative and nonperturbative can lift moduli. W . 1 Requiring the U (1)R charge to vanish implies that b = NF −NC . SU(NC ) Qi a ˜ Qa i 3NC −NF M detM NC ¯ NC 1 1 1 SU(NF )L NF 1 1 NF 1 SU(NF )R 1 ¯ NF 1 ¯ NF 1 U(1)V 1 −1 0 0 0 U(1)A 1 1 2NF 2 2NF U(1)Rcl 1 1 2NC 2 2NF U(1)R NF −NC NF NF −NC NF Λ 0 2− 2NC NF 2(NF − NC ) (5. There is no Coulomb phase so Wα does not appear. In what follows we examine the quantum moduli space for the separate cases: 1 ≤ NF ≤ NC − 1. Rabinovici: Supersymmetric Gauge Theories 5. 2 2 NF = 3NC . has the following form: W = (Λ3NC −NF )a (detM )b c (5.

For NF = NC − 1 there is complete symmetry breaking and one can turn to weak coupling. It gives results for the phase structure in many cases. 6. it has no ground state. The potential for 1 < NF < NC . 1 (5. We will also discuss when these conditions are met. M ij 5. We begin by reviewing the conventional method of integrating out. see [54–62].31) . We will show that under certain circumstances it is possible to derive in a rather straight forward way the potential for light ﬁelds.186 Unity from Duality: Gravity.4 Integrating in This method involves the addition of very massive ﬁelds to known eﬀective actions and extrapolating to the case where the additional degrees of freedom are massless.30) One may now go to NF < NC − 1 by adding masses and integrating out the heavy degrees of freedom. min = (m−1 )i j (Λ3NC −NF detm) NC . This produces. From instanton calculations one calculates that c = 0 and the prepotential for the matter ﬁelds is W ∼ Λ2NC +1 detM · (5. We thus give some of the ﬂavour of this possibility. heavy degrees of freedom are integrated out to obtain an eﬀective potential for the light degrees of freedom. It is rather surprising that anything useful can be learned by this ﬂow in the opposite direction to the usual infra-red. What is the value of c? This is a diﬃcult to calculate directly unless there is complete Brouting. Gauge Theory and Strings V det(M) Fig.

7. Mmin → ∞ as m → 0. It is convenient for general m = 0 to write. One thus obtains: ˜ ˜ ˜ Wd (X. Λd ) ˜ ˜ (5. The eﬀective potential is the Legendre transform ˜ Wu (Φ) = (W (gi ) − where ˜ ∂ W (gi ) = Φi · ∂gi (5. The da dofs will be kept light throughout all the discussion. Berman and E. The potential with ﬁnite masses has a ground state. (Λu is the dynamically generated scale of the theory. Z. M. Λu ) = mM + λZ) M . Λu ) = Wd (X. Λd ) + WI (X. Assume that one is given the eﬀective potential Wu (X.D. One ends with Wd (X. ˜ ˜ Wd (X. M. Λd ). Λu ) = (Wu (X.35) .34) For the case m → ∞ one may tune the scale Λu so as to replace an ap˜ propriate combination of λ.32) Consider next making the microscopic “up” ﬁelds massive and integrating them out retaining only the light degrees of freedom and the couplings m.S. Λu ) describing all the light macroscopic degrees of freedom. Z · (5. m. m. λ ˜ to the macroscopic degrees of freedom containing the removed ﬁelds: M and Z. m and Λu by Λd the scale of the theory of the ˜ remaining light degrees of freedom. λ. Consider a theory containing gauge invariant macroscopic light ﬁelds of the following nature: Fields X: built out of dA degrees of freedom Fields M : built out of ui degrees of freedom Fields Z: built out of both da and ui . Rabinovici: Supersymmetric Gauge Theories V 187 MMin M Fig. Z. λ. The ui and with them the macroscopic ﬁelds M and Z will be considered as heavy in part of the discussion.33) g i Φi ) gi (5. m. λ.

that is. and quark ﬂavors Qi .d f (t) ˜ (5. For the cases corresponding to the colour group SU (2) this indeed turns out to be the case.d + W∆ where Wtree. W∆ (.38) Now one reverses the direction and integrates in.d = (Wtree ) ui (5. λ ˜ .42) where b1 = 6 − 2Na − NF . Writing t as. or m → ∞ . λ.39) All the essential complexities of the ﬂow lie in the term W∆ . The quantum numbers of all relevant ﬁelds and parameters are given by: X λ Λb1 m ˜ W∆ U (1)Q 2 −2 2NF 0 0 U (1)Φ 0 −1 4Na −2 0 U (1)R 0 2 4 − 4Na − 2NF 2 2 (5. which contains Na adjoint ﬁelds Φα with the superpotential: ˜ Wtree = mX + mM + λZ. This can be seen in the following manner.d − Wtree ) m . ˜ (5. given an exact form for Wd (X.NF (NF ≤ 4) for the down theory. Λ) = Wtree. Λd ) + W∆ + Wtree. one resuscitates the full up theory. Λd ) is the exact result for inﬁnitely heavy Ui . W∆ should obeys the constraints: W∆ → 0 . WI = Wtree.40) where X = W W and t being any possible singlet of SU (2NF ) × U (1)Q × U (1)Φ × U (1)R . M. and to partition again. For convenience one writes. . Namely. one brings down inﬁnitely heavy ﬁelds in the adjoint representation.188 Unity from Duality: Gravity. Φ is the adjoint ﬁeld we add. where M = ΦΦ and Z = QΦQ.41) (5. i = 1. t ∼ (Λb1 )a mb X c λd . Λd ) one can obtain: Wu (X. One starts with SU (2) without adjoint ﬁelds (Na = 0). Imagine however that this term would vanish. Λu ) = (Wd (X.. Gauge Theory and Strings where Wd (X. m. In this case the calculations become much simpler.43) . when Λu → 0 .37) has by deﬁnition no scale dependence. (5.36) (5. ˜ (5. Z.

3. W∆ = r(mΛ)b1 ˜ + . Rabinovici: Supersymmetric Gauge Theories It is a singlet provided that b = (2Na + 2 − NF )a .) . 2. (5.45) as well as the fact that in the Higgs phase one can decompose W as: ∞ W (Λb1 ) = n=1 an (Λb1 )n . Recall the constraints W∆ → 0 for m → ∞ and W∆ → 0 for ˜ Λ→0 c = (NF − 4)a . b1 = 4 − NF . Berman and E. m ˜ (5. For NF = 0 and Na = 1 b = 4a leading to ∞ (5. We see this explicitly in the following cases. For Na + 1. Starting from known results for Na = NF = 0 one can now obtain the eﬀective potential for all relevant values of Na and NF .. NF = 1. d = (2NF − 4)a .46) One shows that Wtree.D. 189 (5.44) (5.47) which implies that for all values of Nf the vanishing of W∆ .48) W∆ = a=1 ra (Λb1 )a ma−1 .S. (The role of these elliptic curves in N = 2 theories will be described later. ˜ (5.50) implying that r = 0 as well.49) The constraints imply that r − a indeed vanishes for all a’s.d ∼ 1 m ˜ (5. The equations of motion of these potentials can be arranged in such a manner as to coincide with the singularity equations of the appropriate elliptic curves derived for systems with N = 2 supersymmetry and SU (2) gauge group..

it turned out that m → 0 implied M i j → ∞ thus explicitly lifting the classical moduli space. b˜ (5. (5. Note.52) This has the eﬀect of resolving the singularity in moduli space.6 NF = NC Quantum eﬀects alter the classical constraint to: ˜ detM − B B = Λ2NC . Therefore detM = B B through quantum eﬀects.56) .53) since m cancels. In this case.57) (5. This implies. detM = Λ2NC (5. thus this also holds in the limit m → 0.190 Unity from Duality: Gravity.55) (detM )i+j ij then by demanding that there be no singularities at small M or at large ˜ B implies that all cij must vanish and the hence detM − B B = Λ2NC obeys a nonrenormalisation theorem. B = B = 0 if detm = 0 because all ﬁelds carrying B number are ˜ integrated out. for the case of NF < NC . 1 (5. Writing out an expansion that obeys the R-charge conservation: ˜ (B B)i Λ2NC j ˜ cij detM − B B = Λ2NC 1 + (5. Gauge Theory and Strings 5. 5. mass terms given by: W = tr(mM ) + bB + ˜B. The absence of a singularity means there will not be additional massless particles. mij for matter one obtains. Note. On the other ˜ hand. M i j = (m−1 )i j (Λ3NC −NF detm) NC . There are allowed soft perturbations. R(MIJ ) = 0 for NF = NC . Mi j = (m−1 )i j (detm) NC Λ2 . For NF ≥ NC it is possible to have m → 0 while keeping M i j ﬁxed. In the presence of a mass matrix.54) 1 (5. R(M ) = 0 ⇒ W = 0 .51) Previously.5 Quantum moduli space for NF ≥ NC There is a surviving moduli space.

B = 0 then one is in a Higgs/conﬁning ˜ phase.60) ˜ BB = 0 (5. In both cases the ’t Hooft anomaly conditions [66] are satisﬁed.7 NF = NC + 1 The moduli space remains unchanged. Consider some examples: With the following expectation value. for small M/B/B one is in the conﬁning regime. is it in a Higgs/conﬁnement phase? For ˜ ˜ large. M/B/B one is sitting in the Higgs regime. At the singular point when.S. This is an interesting situation because there is a dogma that as soon as a system has a bound state there will be chiral symmetry breaking [65].61) (5. where because of strong infra-red ﬂuctuations there are n independent ﬁelds even though there is a classical constraint. Berman and E. M = B = B = 0 there is no global symmetry breaking but there is “conﬁnement” with light baryons.62) ˜ BB = 0 (5. This is not a theory of massless gluons but a theory of massless ˜ mesons and baryons. M ij = 0 .63) . ˜ There is a suggestion that in this situation. The classical and quantum moduli ˜ spaces are the same and hence the singularity when M = B = B = 0 remains. B. The analogy is from the nonlinear sigma model.59) which has chiral symmetry and also has conﬁnement.D. 5. B become dynamically independent. then the group is broken to: SU (NF )L × SU (NF )R × UR (1) (5. B.58) What is the physics of this theory. W = ˜ Λ2NC +1 mΛ2NC = · detM detM 191 (5. M. When. M i j = δ i j Λ2 . The eﬀective potential is: Weﬀ = 1 ˜ (M i j Bi B j − detM ) Λ2NC −1 (5. M. the global symmetries are broken to: SU (NF )V × UB (1) × UR (1) and there is chiral symmetry breaking. These will be discussed later. When. however. Global symmetries need to be broken in order to satisfy the modiﬁed constraint equation. Rabinovici: Supersymmetric Gauge Theories One check is to integrate out to give the case NF = NC − 1 yielding. Note that M cannot be taken smaller than Λ.

the diﬀerence between the Higgs and conﬁnement phases is purely quantitative. Gauge Theory and Strings the classical limit is taken by: Λ→0 (5. The phases of super QCD. 8.F Dual Descriptions 3NC 2 Magnetic IR Free. SU (2) × U (1). IR FREE NF 3N C Conformal Two A. which is ZN for SU (N ).192 Unity from Duality: Gravity. This contrasts the situation of pure QCD or super QCD where all the particles are in the adjoint representation which is trivial under the center.64) which in turn imposes the classical constraint. Again the system obeys the ’t Hooft anomaly matching conditions. Is it in a Higgs or conﬁnement phase? Below we . There is no phase boundary. In such a case there is a phase transition and there is a qualitative diﬀerence between the phases. S=1 NC+2 NC+ 1 NC NC-1 No Vacuum 1 0 No Global Chiral Symmetry Breaking Modified Moduli Space !! Confinement < λλ> = 0 Discrete Vacua Fig.8 Higgs and conﬁnement phases This section is discussed in [63]. 5. with matter in a nontrivial representation of the center of the group. So what about the standard model. While one is discussing the conﬁnement phase in supersymmetric gauge theories one should recall that for gauge theories such as SU (N ) Yang Mills.

(l)R . φ(x) = Ω(x) then Bµ = Ω + D µ Ω leading to the Lagrangian. (q)R are SU (2)L singlets. One may choose a gauge. (5. ρ(x) 0 (5.66) (5. ψ2 = . The Higgs picture also contains the operators: ψ1 = φ+ ψi φi ψj ij i . (5.75) Like conﬁnement but with the scale determined by: |φ|.71) ij . W 0 = i Dµ i · |φ| |φ| (5.73) (5. In the Higgs picture. + L = trFµν (B)F µν (B) + ∂µ ρ∂ µ ρ + ρ2 (Bµ B µ )|| + V (ρ2 ).65) (5.74) Unitary gauge is Ω = 1.D. s= 1 2 s=0 s=1 φ+ ψi . Along with. Berman and E. In the Englert picture: s=0 1 2 1 s= 2 s= In the conﬁnement picture.76) . At ﬁnite temperature the two phases are qualitatively indistinguishable. Q(ψ) = 1 1 e Q(φ) = e 2 2 Q(W 0 ) = 0.72) (5.70) (5. Examine the charges of the ﬁelds with respect to the unbroken U (1). i (5.68) (5. |φ| |φ| + φ+ j ˜+ φ Wµ = i Dµ |φ| |φ| ij φ+ φ+ ˜ . φ+ Dµ φ+ ij i j .69) φ+ Dµ φi .S.67) (5. φ+ φi i φi Dµ φj ij ψi φj 193 1 2 1 I= 2 I= I=0 φ → φreal (l)L (q)L (l)R (q)R . Rabinovici: Supersymmetric Gauge Theories present the spectrum in the two pictures.

The appropraite conserved charge is actually: Q = Q + T3 . 5. Is it possible to have strong “weak” interactions at high energies? This is theoretically possible but it is not the course chosen by nature for the standard model. when observed at longer and longer length scales. This is seen empirically by the absence of radial excitations of the Z particles [64]. (5. this will not work for the Georgi-Glashow model where φ is a triplet in SO(3). System I t t t t System II t t t t) • Fig.9 Infra-red duality Two systems are called infra-red dual if. the conﬁned objects have integral charge.194 Unity from Duality: Gravity. Seiberg has observed and has given very strong arguments that the following set of N = 1 supersymmetric gauge theories are pairwise infra-red dual [88]. . Q (e) = 1 and Q (ψφ) = 1 Q (φ+ Dµ φ) = 0 Q (φ + ψ) = 0 Q (φ+ Dµ φ+ ) = 1 Q (φ+ φ) = 0 Q (φDµ φ) = −1. 9.79) (5.77) This matches the charges of operators in the conﬁned picture. Note. Then. they become more and more similar. Gauge Theory and Strings However. Two systems with a diﬀerent ultra-violet behavior ﬂowing to the same infra-red ﬁxed point.80) Q (W 0 ) = 0 Q (v) = 0 Q (W ± ) = ±1 Q (ρ) = 0 (5.78) (5.

being the spin-0 Goldstone Bosons of the spontaneously broken chiral symmetry. Thus the infra-red behavior of the strongly-coupled system is described by infra-red free spin-1 massless ﬁelds (as well as its superpartners). Seiberg’s duality allows for the ﬁrst time to ascribe a similar role to massless infra-red free spin-1 particles. QCD. one can realize a situation where free 1 massless spin. It is infra-red dual to a theory with NF −NC 2 colors and NF ﬂavors and NF color singlets. Seiberg’s work has shown that infra-red free massless spin-1 particles(for example photons in a SUSY system) could be. Nevertheless. For a given NC .2 particles are also the infra-red resolution of another theory. while the dual theory has such a large number of ﬂavors that it becomes infra-red free. that this is the case with Seiberg’s duality. the original theory 2 is strongly coupled in the infra-red. whose gauge group is SU (NC ) and whose ﬂavor group is SU (NF ) × SU (NF ) × U (1). just the infra-red limit of a . By using supersymmetry.S. Note that the rank of the dual pairs is usually different. is expected to be infra-red dual to a theory of massless pions which are all color singlets. QCD whose ultraviolet behavior is described by other particles. for example. under certain circumstances. NC . We have thus relearned that free spin-0 massless particles can actually be the infra-red ashes of a stronglyinteracting theory. the number of ﬂavors. In general. Rabinovici: Supersymmetric Gauge Theories 195 System Gauge Group SU (NC ) SO(NC ) Sp(NC ) #ﬂavors NF NF 2NF Dual System Gauge Group SU (NF − NC ) SO(NF − NC + 4) Sp(NF − NC − 2) #ﬂavor #singlets NF NF 2NF 2 NF 2 NF 2 NF . Massless spin-1 particles play a very special role in our understanding of the basic interactions. if the number of ﬂavors is in the interval NC + 1 < NF < 3NC . For example.D. it has been known for quite a long time that two systems which diﬀer by irrelevant operator have the same infra-red behavior. NF . that is. the N = 1 supersymmetric model with NC colors and NF ﬂavors. for which the infra-red duality holds is always large enough so as to make the entries in the table meaningful. the common wisdom in hadronic physics has already identiﬁed very important cases of infra-red duality. are actually infra-red free in four dimensions. This comes about in the following way: consider. We have no indication whatsoever. Lets explain why this result is so powerful. For a given number of colors. where groups with diﬀerent number of colors are infra-red dual. The pions. Berman and E.

a continuous path in parameter space relates a pair of two disjoint infra-red-dual ﬁeld theories. 10. The infra-red duality relates two disconnected systems.196 Unity from Duality: Gravity. we will need to learn to use some tools of string theory. NF fundamental and antifundamental matter and NF . Consider the example of N = 1 supersymmetric Yang¯ Mills theory with gauge group SU (NC ) and NF . The charges of the matter ﬁelds are given by the table below: SUL (NF ) SUR (NF ) UB (1) Q ˜ Q NF 1 1 ¯F N 1 −1 UR (1) 1− 1− NC NF NC NF · (5. In order to be able to appreciate how that is derived. From the point of view of string theory the two systems are embedded in a larger space of models. We will describe the ingredients of such an embedding [68. Gauge Theory and Strings much more complicated ultraviolet theory. In string theory. Seiberg’s duality has passed a large number of consistency checks under many circumstances. NF fundamental. antifundamental matter. 69] later. First we will describe some more details of the Seiberg infra-red duality in ﬁeld theory. such that a continuous trajectory relates them.81) The infra-red dual is N = 1 supersymmetric Yang-Mills with gauge group 2 ¯ S(NF − NC ) and NF . Fig.

. q mesons in the original SU (NC ) theory? ˜ • the resolution to the previous two points is provided by the interaction term (5. ˜ • The Baryons built from Q. ˜ For the case. What does matter is that the two dual theories share the same global symmetries. For the case NF = NC + 1 then the Baryon of the SU (NC ) ˜ theory becomes the q in the dual theory (which is a singlet in this case). 3NC < NF < 3NC . 2 2 The operator.83) • The dual theories have diﬀerent gauge groups. q . If one regards a gauge symmetry as a redundancy in the description of the theory then this is not important.D. j µ (5. • Where are the q. This looks like there is a solitonic dual for the quarks in this case.82) 0 −N 2 NFNF C · One must also add an interaction term in the dual theory described by: W = 1 i ˜ M ˜ qi q j . Rabinovici: Supersymmetric Gauge Theories gauge singlets. • Note. q mesons. M q q has dimension: ˜ D(M q q ) = 1 + ˜ 3NC <3 NF (5. Q. and for 3 (NF − NC ) < NF < 3(NF − NC ). which implies one may ignore them in the ˜ infra-red. Q have the same charges as those built from q.83). The M ﬁeld in the dual theory ˜ does have the same charges as a meson built from Q. This term breaks the UM (1) symmetry and provides a mass to the q. it is not possible to build a meson out of q. • If M is fundamental there should be an associated UM (1) charge which does not appear in the original SU (NC ) theory. Berman and E. In both dual pictures there is an Infra red ﬁxed point. both are asymptotically free and in the center of moduli space the theories will be a conformal.84) and so it is a relevant operator. Q. The charges are given by: SUL (NF ) q q ˜ M ¯ NF 1 NF SUR (NF ) 1 NF ¯ NF UB (1) NC NF −NC NC − NF −NC 197 UR (1) 1− 1− NC NF NC NF (5. q that has the same ˜ ˜ R-charge as a meson built from Q.S.

SU (NF )2 UR (1) . For the case 3NC < NF < 3NC the two dual theories are both asymptotically free. This will be the subject of the next section. This is the ﬁrst example of a weakly interacting theory with spin one particles that in the infra-red one may view as bound states of the dual theory. Gauge Theory and Strings The checks of the duality are as follows: • They have the same global symmetries. It 2 is symmetric around NF = 2NC . 3 2 3 2 (5. UB (1) UR (1) . Let us examine some of the consequences of this duality. the theory is an infra-red free gauge theory plus 2 free singlets. The ’t Hooft anomaly matching conditions are determined as follows. Adding a mass in one theory is like an Englerting in the other. UR (1) . There was no a priori reason for them to do so. Let us now enrich the structure of the theory by adding Na particles in the adjoint representation. UB (1) . One then calculates their anomalies.85) (5. • There is a construction of the duality by embedding the ﬁeld theory in string theory. • There are the same number of ﬂat directions. Perhaps one can more learn about this system since it is a ﬁxed point under duality. SU (NF )2 UB (1) . In the above example there are anomalies for: SU (NF )3 . The gauge invariant macroscopic moduli would be Trφk . The potential for the scalars. UR (1) . At ﬁrst we will have no matter in the fundamental representation and scalar multiplets which are adjoint valued. Both dual theories must share the same anomalies. • They obey the ’t Hooft anomaly matching conditions. • There is the same reaction to a mass deformation.198 Unity from Duality: Gravity.87) This potential obviously has a ﬂat direction for diagonal φ.86) All these anomalies match between the dual theories. φi is given by: V = ([φ. the supersymmetry is now increased to . (5. Consider the non generic example of NC = 2 and Na = 1. • It is a Z2 operation. At the origin of moduli space one may have obtained a new conformal theory.this will be discussed later. UB (1) UR (1). One takes the global symmetries in the theory and then make them local symmetries. For NC + 2 ≤ NF ≤ 3 NC . The panorama of these structures is given in Figure 8. φ])2 .

S. m → −e.91) (5. Λ. This generalizes to a U (1) symmetry by deﬁning: E + iB .89) τ → − . Λ)Wα W α ). as follows: 1 (5. new particles will become massless. Classically.90) . The exact quantum potential vanishes in this case [48. B → −E . Everywhere else the theory is in the Coulomb phase. τ → τ + 1. At the special points in the moduli space. one would expect that the strong infra-red ﬂuctuations would wash away the expectation value for Trφ2 and the theory would be conﬁning. This theory has an SL(2. e + im. This is illustrated in the Figure 11. deﬁned by 5.D. L= d2 θIm(τeﬀ (trφ2 . e → m . We will now examine the eﬀective theory at a generic point in moduli space where the theory is broken down to U (1). The surprising thing is that when SU (2) breaks down to U (1). Recall the usual electromagnetic duality for Maxwell theory in the presence of charged matter is: E → B . 49]. The Lagrangian is given by. Z) act on τ . there are only two special points in moduli space and even there the theory is only on the verge of conﬁnement. g. τ This is a generalization of the usual U (1) duality that occurs with electromagnetism to the case of a complex coupling. One would expect a singularity at Trφ2 = 0. There is a single complex modulus. Rabinovici: Supersymmetric Gauge Theories 199 Fig. SU (2) is broken to U (1) for Trφ2 = 0.2.88) The τeﬀ is the eﬀective complex coupling which is a function of the modulus. (5. When Trφ2 is of order Λ or smaller. N = 2. The generators of the SL(2. (5. Z) duality symmetry. naive quantum and exact quantum moduli spaces. Naively. trφ2 . Trφ2 . Berman and E. the original couplings and the scale. 11. The classical. because of the very strong constraints that supersymmetry imposes on the system. one would expect the following.

(5. This occurs for deﬁnite values of u in the moduli space. 62]. Again the classical moduli space is singular at times. For condensation to occur the monopoles should become tachyonic indicating an instability that produces a condensation. NC . 2 τeﬀ = i∞ . a.95) .93) determines a torus. geﬀ = 0 (5.. These new massless particles are monopoles or dyons. One can push this to conﬁnement by adding a mass term: mTrφ2 . 59. there is no perturbative or nonperturbative corrections. The following complex equation.200 Unity from Duality: Gravity. How does one ﬁnd τ as a function of the u? There is a great deal of literature on the subject here we will just sketch the ideas [48–50]. The complex structure of the torus. . k = 2. When y(x) and y (x) vanish for the same value of x then τ is singular. (5. y 2 = ax3 + bx2 + cx + d (5. or generally for SU (NC ) ˜ the term: δW = gk uk . The eﬀective prepotential is now: W = M (uk )q q + gk uk ˜ then ∂W ∂W = 0 ⇒ M ( uk ) = 0 . τtorus will be identiﬁed with the complex coupling τeﬀ . couplings and scale and so will implicitly determine τtorus . b.97) =0 . The monopoles are massless but they have not condensed [54–57. e + im → exp(iα)(e + im). ˜ ∂uk ∂(q q ) ˜ (5. The theory is on the verge of conﬁnement. Gauge Theory and Strings The duality symmetry now acts by: E + iB → exp(iα)(E + iB) . c. This reﬂects the presence of massless charged objects. Therefore. d are holomorphic functions of the moduli.. 60.92) Previously for the SU (2) case the moduli were given by u = Trφ2 for SU (NC ) the moduli are given by uk = Trφk .94) and the eﬀective coupling vanishes. For N = 2 supersymmetry that is the best one can do.96) (5. ∂uk M ( uk ) q q = −gk .

The equations of motion from minimizing the superpotential are: ∂W2. Z) = Mαβ + Tr2NF (Zα X −1 Zβ X −1 ) which we will use to write the prepotential as follows. WNF .. NC = 2 with arbitrary NF and Na . α. We deﬁne the quantity. a.. NF = 2. Take the case Na = 1. The ﬁelds that are the moduli in the system are: XIJ Mαβ Zij = = = a b ab Qi Qj aa aa ab a b bb φα φβ bb (5. 2 (5.100) (5.106) = 2Λ−1 (P f X) 2 = R−1 (X −1 − 8Γ−1 X −1 (ZX −1 )2 ) = 4R−1 Γ−1 X −1 ZX −1 (5..109) 1 Γ. This is achieved here by using the integrating in method discussed earlier. Rabinovici: Supersymmetric Gauge Theories Since generically. ∂uk M ( uk ) = 0 201 (5. Z) = (b1 − 4) Λ−b1 P f X(det(Γαβ ))2 1 +TrNa mM Tr2NF mX ˜ 2 1 + √ Tr2NF λα Zα . b = 1. 2.98) then there will be condensation.107) (5.1 = = =0 ∂M ∂X ∂Z which imply: m ˜ m 1 √ λ 2 where R−1 ≡ Λ− (P f X) 2 Γ .104) (5. X.110) .108) (5. 2 1 4−b1 (5..101) Qa φa b Qb i α j where Q are fundamental and φ are adjoint ﬁelds. 2NF.Na (M. Na . Γαβ (M.1 ∂W2.1 ∂W2. We now describe how the complex elliptic curve arises using more physical terms.S.102) (5. X. X ≡ 1 1 (5.103) (5. Consider the case. j = 1.99) (5.D. Berman and E. β = 1.105) This respects the necessary symmetries and can be checked semiclassically. i.

it is claimed that the theory is described by a non-trivial conformal ﬁeld theory [73]. 2 5. There are several motivations for reaching this conclusion. (5.111) (5.202 Unity from Duality: Gravity. y 2 = x3 + ax2 + bx + c. the exact β function: β(g) = − g 3 3NC − NF + NF γ(g 2 ) 2 16π 2 1 − g 2 NC 8π (5.115) (5. α≡ 16 (5. • These techniques may be extended to obtain curves for other more complicated groups. We now rexamine some special properties of the region 3 NC < NF < 3NC . This modular parameter will then be the eﬀective coupling of the theory. Examine for example. are degenerate vacua which are possibly non-local with respect to each other. • As you move in moduli space monopoles turn smoothly into dyons and electric charge. These are Argyres Douglas points [71]. This is an indication of the Higgs/conﬁnement complementarity. at the center of moduli space when all 2 expectation values vanish.112) Taking y and y to vanish we can compare with the elliptic cure. Gauge Theory and Strings The following equations are then obeyed: X 3 − M X 2 + bX − 1 (c − 8M )˜ = 0 c 128 X 2 − 2M X + b = 0. µ ≡ λ−1 m. One can therefore identify the parameters as: a = −M c= α (2M + Tr(µ2 )) 8 b=− α Λ2 + Pfm 4 4 Λ4 .116) .113) Identifying the modular parameter of the torus from the elliptic equation involves standard techniques in algebraic geometry.114) (5. Some comments: • Some points in moduli space when 2 + NF = 4.10 Superconformal invariance in d = 4 For the case of 3 NC < NF < 3NC .

75].119) ˜ D(QQ) = 2 + γ = 3 NF − NC · NF (5. The product contains a superposition of operators with diﬀerent dimensions. In d = 2 the conformal group is inﬁnite dimensional and so provides very powerful constraints on the theory.122) Generally. For chiral .D. Nevertheless there are still interesting properties of d = 4 conformal theories arising from the conformal invariance.S. NC are both taken to be large but their ratio is kept ﬁxed then the ﬁxed point of the β function. 8π 2 NC 203 (5. (5. is at g ∗2 = 1 8π 2 . The bound is saturated for (anti-)chiral ﬁelds. Since the coupling at the ﬁxed point is proportional to << 1. In d = 4 the conformal group is ﬁnite dimensional and so the conformal symmetry does not constrain the theory in the same way [74. One can prove that the scaling dimension of a ﬁeld is bounded by its R-charge as follows: D≥ 3 |R| 2 (5.118) NF where = 3 − NC << 1. β(g ∗ ) = 0.120) This observatoin will lead to interesting conclusions.122). Berman and E.121) where D is the scaling dimension and R is the R-charge. Consider the operator product: O1 (x1 )O2 (x2 ) = i i O12 f i (x1 − x2 ).117) If one now considers a limit where NF . NC 3 (5. both the existence of a ﬁxed point and the perturbative evaluation for γ is justiﬁed. Rabinovici: Supersymmetric Gauge Theories where γ(g 2 ) is the anomalous dimension given perturbatively by: γ(g 2 ) = − 2 g 2 NC − 1 + O(g 4 ) . the dimension of the product of the operator appearing on the righthand side of equation (5.122) is not a sum of the dimensions of the two operators appearing on the left hand side of equation (5. The anomalous dimension at the ﬁxed point is: γ(g ∗ ) = 1 − ˜ The dimension of the QQ is 3NC · NF (5.

The closure property of the chiral ﬁelds under the operator product expansion leads to the name “chiral ring”. This will be elaborated later. The dimension of a chiral ﬁled is: j D(O) ≥ j + ˜ + 1 j (5. D(I) = 0 .123) i and f i (x1 − x2 ) are thus all constants.204 Unity from Duality: Gravity. For a free ﬁeld the bound is saturated. There are reasons to expect that at the ﬁxed point. where I is the identity operator and O is an operator = I. ˜ respectively.128) . K. the operator QQ becomes free. This NC + 1 < NF < 2 NC it appears that D(QQ) is an indication that one is using the wrong degrees of freedom and a dual description is required. Since the R-charge is additive and chiral ﬁelds saturate the bound (5.121) i R(O12 ) = R(O1 (x1 )) + R(O2 (x2 )) ⇒ D(O1 O2 ) = D(O1 ) + D(O2 ) (5. see equation (5. SU (3) theories without matter or SU (2) theories with matter [76]. j Unitary chiral primaries obey: j ˜ = 0. ˜ charge as well as D the dimension.126) ˜ ˜ When NF = 3 NC . Non-trivial superconformal N = 2 theories in d = 4 occur for N = 2. For 2 3 ˜ < 1 which is forbidden. The key point is that a non-trivial conformal theory with vector ﬁelds contains both massless electric and magnetic excitations (these are mutually non-local). the infra-red nonanomalous R-charge equals the non-anomalous R-charge of the ultra violet. Gauge Theory and Strings ﬁelds the situation is simpler.120): ˜ D(QQ) = NF − NC 3 ˜ · R(QQ) = 3 2 NF 3NC (NF − NC ) · 2NF (5. The deﬁnition of a primary state is that it is annihilated by the generator of special conformal transformations. The descendents are obtained by acting on primary states with momentum operators. since D(QQ) = 1.123) O12 is also a chiral operator. (5.124) The dimension of the Baryon and anti-Baryon are: ˜ D(B) = D(B) = (5. An operator with non-zero spin will be carry j. ˜ At the ﬁxed point of the dimension of QQ is given by. The Lorentz group dej composes into SU (2)L ⊕ SU (2)R with charges j. D(O) ≥ 1.125) For unitary representations (of spin = 0) ﬁelds. By (5.127) (5. D(O) = 1.

dotted indices are SU (2)R i indices and 2 M 12 = J 3 .132) These states are the decendents of F ± . Jm ≡ J+ − J− (5. j For a free ﬁeld: D(O) = j + ˜ + 1. (5. ||J ± | = 2(D − 2). (5.133) where the undotted indices are SU (2)L indices.135) For D = 2.131) One can show that there are states associated with the conserved currents: ± ± Jµ = ∂ ν Fµν . (5.137) (5.134) and that K annihilates primary states. ˙ (5. the Hermitian conjugate relation.130) This generalizes the previous results for scalar chiral ﬁelds.D. these are null states and F ± is free. (5. called the Argyres Douglas point [71]. The norm is then calculated using the following: the commutator of the d = 4 conformal algebra. At the point in moduli space (discussed previously) where there are both electric and magnetic charges. ˙ (P αα )+ = − αβ αβ ˙ ˙ K ββ . F ± . Rabinovici: Supersymmetric Gauge Theories as compared with a non-chiral operator: D(O) ≥ j + ˜ + 2. . F = F + + F −. F ± = F ± ∗F. After some algebra this produces.S. Je ≡ J+ + J− . obtained by applying the momentum operator P on the chiral ﬁeld. ˙ [P αα .136) are non-vanishing [76]. K β β ] = ˙ i αβ M 2 αβ ˙ ˙ ¯˙˙ + M αβ αβ +D αβ αβ ˙ ˙ . Berman and E. Fµν decomposes into self-dual and anti self-dual parts that form irreducible representations of SU (2)L ⊕ SU (2)R . j 205 (5. then ||J ± | > 0 and therefore both. If F is not free and D > 2.129) (5.

144) = −f (g(µ))3T (R)γ(φ) .140) where dw is the engineering dimension of the composite perturbing term.142) T (Rk )) + k T (Rk )γ(φk )) (5. consider. For a general hφ1 . d(φk ) is the engineering dimension of the ﬁeld. k 1 3 γ(φk ) = γ 2 2 (5.139) When h = g. g being the gauge coupling. this is actually a deformed N = 4 theory. the theory is conformal when NF = 2NC and the one loop β function vanishes. X3 : Lint = hX1 X2 X3 . φk and γ(φk ) is the associated anomalous dimension of the ﬁeld φk . Other conformal theories occur for N = 1 supersymmetry when Na = 3 and the couplings are appropriately tuned. X2 .206 Unity from Duality: Gravity. One can deform N = 4 theories with marginal operators such that the global symmetries are broken but the theory remains conformal. By symmetry all ﬁelds have the same anomalous dimension and so γ(φk ) is independent of k and so is denoted as simply γ(φ). k=1 d(φk ) = 3 . In the case at hand 3 dw = 3 .φn perturbation one has n βh = h(µ) −dw + k=1 1 d (φk ) + γ(φk ) 2 (5. this is known to be an exact result [77].. the theory has full N = 4 supersymmetry. For N = 2 supersymmetry. Hence.138) where Oi are the set of operators with dimension 4. (5. L = L0 + i gi O i (5. Consider N = 1 with Na = 3 and an interaction for the adjoint ﬁelds X1 .141) and therefore 3 βh = h(µ) γ(φ) 2 and βg = −f (g(µ))((3C2 (G) − k (5. Gauge Theory and Strings this condition is met and it is possible to have a non-trivial conformal ﬁeld theory with spin one particles.143) (5.

Z) duality [80] that identiﬁes theories with coupling τ given by equation (5.) This ﬁxed line is infra-red stable. Therefore. c. one can ask if the N = 4 theory has any discrete symmetries that relate theories with diﬀerent moduli. There is a great deal of evidence that N = 4 theories possess an SL(2. which is non vanishing.145) To conclude. 82]. C2 (G) is the second Casimir of the group G and T (Rk ) is associated with the representation of φ. Rabinovici: Supersymmetric Gauge Theories 207 h g Fig. This means there is a ﬁxed line. 6 Comments on vacuum energies in scale invariant theories The puzzle to be addressed in this section is that of the cosmological constant problem [81. cτ + d ad − bc = 1 . b.2): τ→ aτ + b . where g = h and the supersymmetry is enhanced to N = 4.S. At the ﬁxed line. a. (5.D. The answer is yes [79]. 12. RG ﬂow to the ﬁxed line. (This is diﬀerent from the more generic situation where the β functions are not related and there are isolated ﬁxed points. supersymmetric gauge theories have a very rich phase structure and many outstanding dynamical issues can be discussed reliably in the supersymmetric arena that are hard to address elsewhere. Given that there is an eﬀective scale below which the physics is known then (such . Berman and E. The problem was originally stated as follows. rather than a ﬁxed point. d ∈ Z. both βh and βg vanish if γ vanishes. Also the relations between relevant deformations follow a similar pattern [78].

g6 is a modulus. Gauge Theory and Strings as QED or the standard model) integrating out the more energetic degrees of freedom above this scale leads to a vacuum energy that is proportional to the cut oﬀ.84] and also [78.208 Unity from Duality: Gravity. The consequence of scale invariance is a zero vacuum energy whether or not scale invariance is spontaneously broken. βg6 = 0 (6. Giving an expectation value to a ﬁeld will break the scale invariance spontaneously as well as generically breaking the gauge symmetry down to U (1)r where r is the rank of the unbroken gauge group. recall the example discussed in quantum mechanics where the action scale invariance and supersymmetry were spontaneously broken but the system had no ground state. L = ∂µ φ · ∂ µ φ − g6 |φ|2 3 (6. the vacuum energy calculated in such a manner however would give a cosmological constant many orders of magnitude above the observed value.2) where the ﬁelds φ are in the vector representation of O(N ). The presence of a ground state is crucial. global supersymmetry can’t be spontaneously broken whether or not scale invariance is spontaneoulsy broken.4) . When one writes down a low energy eﬀective action for a spontaneously broken theory one should respect all the symmetries that appear in the original action. Thus provided there is a translationally invariant ground state. One should observe some important caveats in this argument [83. If the original theory is scale invariant then the eﬀective action of the spontaneously broken theory should reﬂect this symmetry.85]. σ = |φ|2 (6. As an example consdier N = 4 super Yang-Mills in four dimensions. An eﬀective potential can be written down for the O(N ) invariant ﬁeld. The vacuum energy remains zero. The next example is the O(N ) model in three dimensions. described by the Lagrangian. N → ∞. In the limit.3) 1 ( N corrections break the conformality). An analysis of the spectrum shows that a gauge singlet particle emerges. which is the dilaton (the Goldstone Boson associated with spontaneously broken scale invariance). The potential is given by: V = [φ. This result is exact to all orders in perturbation theory and non-perturbatively.1) thus the theory has ﬂat directions. φ]2 (6. For either choice of the above physics.

with Lagrangian: L = ∂µ φ1 · ∂ µ φ1 + ∂µ φ2 · ∂ µ φ2 − λ6. The ﬁnal example will be the O(N ) × O(N ) model with two ﬁelds in the vector representation of O(N ). Rabinovici: Supersymmetric Gauge Theories the eﬀective potential is: V (σ) = f (g)|σ|3 . There are now two possible scales. Assume a hierarchy of scales where the scale invariance is broken at a scale much above the scale at which the O(N ) symmetries are broken. Berman and E. The vacuum energy remains zero however and not proportional to σ 3 as is the naive expectation. |φ2 |2 (6. summarized in the table below: f (g) > 0 f (g) = 0 f (g) = 0 σ 0 σ =0 σ =0 S.5) (6. one associated to the break down of a global symmetry and another with the break down of scale invariance.2 |φ1 |2 −λ0. One has the following possibilities. One would have argued that one would have had a low energy eﬀective Lagrangian for the pions and dilaton . masses V No 0 0 No 0 0 Yes σ .6 |φ2 |2 2 3 (6.S. in all cases the vacuum energy vanishes. The possibilities are summarized by the table below: O(N ) + − + − O(N ) + + − − scale massless + all − (N − 1)π s.0 0 209 (6. σ 0 σ 0. (For f (g) < 0 the theory is unstable). D − (N − 1)π s.0 |φ1 |2 −λ4. Note. the vacuum energy always vanishes. one could write down the eﬀective theory for energy scales below σ and integrate out the physics above that scale. D − 2(N − 1)π s. (6.9) Again. D massive V none 0 N.7) 2 |φ2 |2 − λ2. To summarize.B. in the situation where σ = 0 and the scale invariance is spontaneously broken.B. σ 0 N.D. indicates spontaneous symmetry breaking of scale invariance and V is the vacuum energy.6) where S.4 |φ1 |2 .8) 3 Again the β functions vanish in the strict N → ∞ limit.

Our tools for this project will be. We obtain them as the low energy limit of various string backgrounds. 7. It was proposed that the underlying physics of nature is scale invariant and the scale invariance is only removed by spontaneous symmetry breaking [83.. from which this “comics illustrated” is derived is yet to be written.1 Branes in string theory Branes are extended object solutions which emerge non perturbatively in string theory in a very similar way that solitons emerge in ﬁeld theory. In the following discussion we will have as our goal to: • Construct: – a D = 4 dimensional eﬀective theory. – with U (Nc ) gauge symmetry. – with SU (NF ) × SU (NF ) × U (1) global symmetry. The hope is that this dilaton may somehow be given a mass with a further independent (Higgs like?) mechanism. 85]. The total energy of the universe can be augmented even in a scale invariant theory. comic strips [95]. The unabridged original novel.210 Unity from Duality: Gravity. 94]. 84] and also [78. the vacuum energy remains zero. The scaﬀolding for this construction will be extended objects called branes. • Identify dualities in a pedestrian way. (The data in 2001 seems to indicate the presence of a small cosmological constant. Gauge Theory and Strings with a vacuum energy given by the scale at which the global symmetry is broken.. Many properties of gauge theories are obtained in this fashion [93. If there are many of them they may mimic a very small cosmological constant [87]. This is not true. This would explain the vanishing of the cosmological constant. – with N = 1 supersymmetry (SUSY). Such theories may have many superselection sectors (such as magnetic monopoles in N = 4 super Yang-Mills). In each sector the total energy will diﬀer from zero by the energy of these particles. 7 Supersymmetric gauge theories and string theory We now view the supersymmetric gauge theories from a diﬀerent point of view.) A key problem with this scenario however is that we do not observe a dilaton in nature that would be expected if scale invariance is spontaneously broken. Magnetic monopoles and vortices are examples of solitonic conﬁgurations .

at this stage. Φ. These excitations are often very heavy and have little direct impact on the low-energy dynamics. In lattice gauge theories. there exist solitons.. In the RR sector. respectively. Gravitons and the other particles in the NS-NS sector have a perturbative string realization. Bµν . such as conﬁnement.. the regulator will also learn how to behave from the very theory it regulates. carrying two and four indices. .. we ﬁrst review the spectrum of massless particles in various string theories [97].S. . µ = 1. the spectrum consists of a graviton. In the closed bosonic string theory. Bµν . . discovered in supergravity.. whose tension 1 T is: TNS−NS = g2 l6 . such regulators have granted a very rapid access to identifying non-perturbative features. respectively. duality relations in ten dimensions between the electric ﬁeld Eµ1 . of course. To appreciate that. It.µ8−p−1 . and the electric ﬁeld Eµ1 . The RR sector consists of vector potentials which are 0-forms. consisting of one-form and three-form vector potentials. Bµν . and a dilaton. which seems to be engraved in at least part of our community.. Rabinovici: Supersymmetric Gauge Theories 211 in gauge theories. In string theory. . However. denoted by Gµν . What about the objects in the RR sector? First we note. s s These solitons spread over ﬁve space and one time dimensions. To each of them is associated an “electric ﬁeld”.D. 2. as well as that of other interesting topological excitations. there existence answers a yearning to search for more than meets the eye (or the equations). Berman and E.2 Branes in IIA and IIB string theories For some years it has been known that solitons exist in the low-energy effective theory of superstring theory/supergravity [96]. The bosonic sector of type-IIA theory consists of two sectors.µp+1 . In the context we will discuss here. Φ. The massless spectrum of the open bosonic string consists of photons Aµ . 24. In addition. 2-forms. in which they can condense and thus take over the control and drive the infra-red dynamics. the branes will basically serve as regulators for some ﬁeld theories. playing the same role as string sizes and lattice cut-oﬀs. The ﬁrst is called NS-NS sector and consists of the same spectrum as that of the bosonic closed string theory. derived from the p-form A. In type-IIB the NS-NS sector also consists of Gµν . The other sector is called the RR sector. The exact background corresponding to this conﬁguration is not yet known as the dilaton background seem to contain singularities. derived from the (6-p)-form A lead to further vector potentials. an anti-symmetric tensor. . that in the NS-NS sector. . Φ.. 7. namely Gµν . there are circumstances. where gs is the string coupling and ls is the string scale. A yearning. and 4-forms. in strong-coupling approximations. What are they good for? First. denoted by Aµ and Aµν . a 5-form in type-IIA and 4-forms and 6-forms in RR sector of type-IIB. there are . Similar consequences will occur here. also answers positively the important question of the existence of a magnetic monopole.

respectively. The translational non-invariance of the Dp-brane . . One may inquire as to the eﬀective theory on the six-dimensional world volume of the NS-NS soliton. They are of the form: L QL + R QR (7. 13). Qβ j } = α {Qi . These are called D-branes. The coordinates x0 . will survive intact. Gauge Theory and Strings solitons as well [92]. The supersymmetry algebra: ¯ {Qi . Γ5 (7. . ±Γ0 . existing in ten dimensions. . Qβ j } = 0 (7. It turns out that. whose extension is in p+1 space-time dimensions breaks translational invariance in 9-p directions. obeying appropriate Dirichlet boundary conditions. The p + 1 dimensional solitons have a tension which is: TRR = g l1 . x9 are ﬁxed. . x1 . xp are unconstrained and span the brane’s space-time p + 1 dimensional world volume. spin-0 particles. The solitonic sector of closed string theory contains conﬁgurations of open strings. It is easier to answer the second question. The massless sector on the NS5 in the type-IIA theory is a chiral system consisting of a self-dual anti-symmetric vector potential.. .212 Unity from Duality: Gravity. In the presence of the NS5 conﬁguration sixteen (half) of the SUSY charges survive. Let us return now to the RR sector. One may also inquire what are the supersymmetry properties of that theory. It p+1 s s turns out that open strings.3) where the sign ± corresponds to type-IIA and type-IIB. in ﬂat space-time. A soliton. and the Γ-matrices are respective ten-dimensional Dirac Γ-matrices. The D-brane solitons in this sector are denoted Dp. . The D stands for Dirichlet. Qj } = α β µ 2σαβ Pµ δ ij ¯ ¯ { Qα i .. may end on these branes. ﬁve scalars. . and their supersymmetric partners.1) contains on its right-hand side the generators of translation. Thus it is not clear a priori if any of the supersymmetric generators. .2) where L and R obey the constraints: L R = = Γ0 . Γ5 L R. residing in the left-hand side of the algebra. The answer to the ﬁrst question is more complicated. The other coordinates xp+1 . The massless sector on the NS5 in the type-IIB theory is a non-chiral system consisting of spin-1 particles. where p denotes the number of spatial dimensions of the branes world volume (the spatial volume in which the brane extends). the NS5 brane. . and their supersymmetric partners. the system has 32 SUSY charges. This is our ﬁrst comic strip (Fig.

1 superpartners. and their spin. The next comic strip describes this conﬁguration for NC = 4 (Fig. unaltering the large distance geometry.9 Fig.. Thus GN s “M ” vanishes at weak coupling for Dp branes.1.p T % open strings end on brane 213 −→ xp+1. spin-1 massless particles. there are sixteen surviving supercharges. 13. Berman and E.. Γp R.2)) obey the constraint L = Γ0 . . . . For Dp-branes R and L in the surviving supersymmetric charge (Eq. It can be shown to contain. and the gauge symmetry has increased to U (1)NC . in addition to the sixteen conserved supercharges.3 The eﬀective ﬁeld theory on branes The eﬀective theory on the Dp-brane can be identiﬁed in more detail in this case.. . one can construct a conﬁguration containing NC parallel Dp-branes.S. Newton’s constant..4) Thus. Similarly. for any p. 14). applying naively Newton’s law. Rabinovici: Supersymmetric Gauge Theories D-brane $ x0.. “M ” is 1 1 proportional to gs for Dp branes and to g2 for the NS5 brane.. . A Dp brane with an open string ending on it.D. This theory still contains sixteen conserved supercharges. The eﬀec2 tive theory is invariant under local U (1) gauge transformations (Fig. (7. one can estimate the eﬀective gravitational 2 coupling GN “M ”. GN . Note that. (7. is proportional to gs . 15). 7. reduces the number of SUSY generators by half for any value of p. 9-p massless scalars (corresponding to the Goldstone Bosons resulting from the spontaneous breaking of some of the translational invariances).

2 s where xi represents the value of the coordinates at which the brane is set. the mass of a particle has a very simple geometrical meaning. The eﬀective ﬁeld theory describing this D4 brane conﬁguration is a U (4) gauge theory. 14.214 Unity from Duality: Gravity. For example. A parameter in ﬁeld theory. 15. will lead to the emergence of massless . A massless state propagating on a Dp brane. Imagine now bringing the parallel branes together (Fig. a string stretched between the i-th and j-th brane represents a particle which has a mass mij = l1 |xi − xj |. This. 16). ¨ © ¨ © ¦¥ §¤ U (1)4 ¨ © ¦¥ ¨ © ¦¥ Fig. spontaneously broken down to U (1)4 . Gauge Theory and Strings ¨ © Fig. Four massless states (corresponding to open strings ending on the same brane) and six massive states (corresponding to open strings ending on diﬀerent branes) are shown. It turns out that the masses of the particles can be directly associated with the minimal distance between the end points of the strings stretched between diﬀerent branes. according to the above relation.

In fact. The system has indeed the same particle content and gauge symmetries as can be inferred from Figure 15. The minimal gauge symmetry for NC Dp-branes is U (1)NC . The sixteen massless states formed are not shown in the ﬁgure.D. The classical potential is ﬂat and allows for an inﬁnite set of vacua. This has a correspondence in SUSY ﬁeld theory. Fig.p+1 is the eﬀective Yang-Mills coupling in p + 1 dimensions. ˜ ˜ (7. Moreover. in a N = 1 supersymmetric gauge theory. only such residual gauge symmetry groups that preserve the rank are allowed according to the comic strip. containing two multiplets with opposite electric charges. X J ]2 . (7.7) 1 2 gYM. one can show that these massless particles can enhance the gauge symmetry all the way from U (1)4 to U (4). X I ] (7.5) 4 µ ls ls gYM. The maximal gauge symmetry is U (NC ).S. For example. parameterized by those expectation values of the scalar ﬁelds X I for which the potential term vanishes.6) X I is a scalar ﬁeld in the adjoint representation of the gauge group and 2 gYM. Indeed. the potential is given by: V = (q † q − q † q )2 .p+1 2 TrFµν + 1 1 D X I D µ XI + 8 2 Tr[X I . the Higgs mechanism in the presence of scalars in the adjoint representation is known to conserve the rank of the group.5) is generic for supersymmetric theories. The Lagrangian describing the bosonic sector of a supersymmetric gauge theory with sixteen supercharges is of the form: L= where Dµ X I = ∂µ X I − i[Aµ . Rabinovici: Supersymmetric Gauge Theories 215 particles. The form of the potential term appearing in equation (7. piled on top of each other.p+1 . Four parallel D-branes. Berman and E. it is proportional to gs . 16.

The vacua of the system consist of those expectation values for which q = q = 0. The Coulomb phase is that phase of gauge theory for which the force between both electric and magnetic charges is a Coulomb force. In each of these vacua. For some theories with a small number of SUSY charges. 7. the separations between the branes should vanish themselves in that limit. namely the brane conﬁguration leading to an eﬀective D = 4. Xii on the right-hand side of the equation denotes the component of the Higgs ﬁeld in the i-th element of the Cartan subalgebra of U (NC ).4 Eﬀective D = 4 dimensional systems with N = 2 supersymmetry To obtain an eﬀective D = 4 description one can either set up a single D3-brane conﬁguration in type IIB string theory or build a more complicated conﬁguration in type IIA string theory. N = 1 supersymmetrical U (NC ) gauge theory containing in addition matter ﬁelds. the system will have ˜ massless scalar excitations. For theories with sixteen SUSY charges. Returning to the conﬁguration of NC Dp-branes. (7. . This is again apparent from the geometry of the open strings ending on the brane conﬁguration.216 Unity from Duality: Gravity.9) is just the usual mass obtained by the Higgs mechanism. these separation should be sub-stringy. It turns out that for our goals .8) xi on the left-hand side of the equation denotes the location of the i-th brane. one cannot resolve the NC diﬀerent world volumes. In this limit. NC ). To keep ﬁxed the mass of these “W ” particles in the limit of the decoupling of the string states (ls → 0). . the potential remains ﬂat also nonperturbatively. The expectation values of the Higgs ﬁelds in the adjoint representation can be shown to have themselves a very transparent geometrical meaning: xi = Xii (i = 1. In this context. 52]. the mass formula mij = 1 |x − xj | 2 i ls (7. Gauge Theory and Strings It is a property of supersymmetric theories that the ﬂat potential is retained to all orders in perturbation theory. The number of massless particles is at most NC (in complex notation). . . we have seen that U (1) symmetries remain unbroken and therefore the theory is said to be in the Coulomb phase. We are by now well on our way to obtain that brane conﬁgurations will help accomplish our goals. that is. so the theory is perceived as a U (1)NC gauge theory on a single p + 1 dimensional world volume. this result is modiﬁed non-perturbatively [15. They are denoted moduli and their number is called the dimension of moduli space.

is eﬀectively four dimensional. Establishing that the D1 may end on a D3-brane. The world-volume of the D3-brane is chosen to extend in the x0 . The content of the ﬁeld theory will depend on the branes “?” on which the D4 brane ends. The ﬁeld . x7 .D. In this conﬁguration a D4-brane (whose world volume is 5 dimensional and of the type M 3. One constructs a brane conﬁguration which has a world volume of the form M 3. x6 . x8 . In the ﬁrst example we will show that a D4 is allowed to end on a D6-brane. A fundamental string (F1) can by deﬁnition end on any Dp-brane. This is realized by the following conﬁguration: Fig. that is the vertex appearing in Figure 18 is allowed. The candidates for “heavier” such branes would be either NS5-branes or D6-branes. that is the ﬁeld theory at energy scales much smaller than 1/∆x6 . in an N = 4 supersymmetric gauge theory in D = 4 dimensions the gauge coupling constant g is a real parameter. 17. the eﬀective ﬁeld theory. Performing what is called an S-duality transformation validates that also a D1brane may end on a D3-brane in type IIB theory.and T -dualities.S. S-duality is a symmetry which is familiar already in some ﬁeld theories. we pause now to brieﬂy discuss several types of useful discrete symmetries in string theory. For example. Rabinovici: Supersymmetric Gauge Theories 217 the latter is more useful.1 × I[∆x6 ] ) the two branes between which the D4-brane is suspended would be chosen such that the eﬀective ﬁeld theory has U (NC ) local gauge symmetry and N = 1 supersymmetry. The eﬀective ﬁeld theory on the D4 brane will be four-dimensional for 1 energies much smaller than ∆x6 . that is on which branes are the D4-branes actually allowed to end. For either choice. called S. in particular in type IIB string theory it can end on a D3-brane. Before analyzing the various eﬀective theories resulting from the diﬀerent choices of the branes on which the D4-brane ends we discuss which are the allowed “vertices”. Berman and E. x9 directions and that of the D1-brane extends in x0 .1 × I[∆x6 ] where M 3.1 is the four-dimensional Minkowski space-time and I[∆x6 ] is an interval of length ∆x6 .

In particular. It describes those excitations in which the closed bosonic string extends and winds around the compact dimensions m times. There is evidence that the theory with coupling g is isomorphic to the theory with coupling 1/g. theory is ﬁnite. similarly in string theory diﬀerent types of branes are interchanged under S-duality. In its implementation in ﬁeld theory electric and magnetic excitations were interchanged. All in all. the gauge coupling does not run under the renormalization group and thus diﬀerent values of g correspond to diﬀerent theories. It is the extended nature of the string which leads to this result. R2 (7. Gauge Theory and Strings D4 D6 Fig. has a generalization involving also the value of the θ parameter in ﬁeld theory and an additional corresponding ﬁeld in string theory. T -duality can . For a small radius R these are very low energy excitations. Type IIB string theory has similar properties with the string coupling playing the role of the gauge coupling. n2 The term R2 is not particular to string theory. it describes also a point particle in a Kaluza-Klein compactiﬁcation. The mass M of the particles depends on the compactiﬁcation radius through the formula: M2 = n2 + m2 R 2 . An F1 is interchanged with a D1. is identical to another bosonic string theory whose compact dimension in string units is of radius 1/R. an interchange of n and m simultaneously with an interchange of R and 1/R in equation (7. We have used some of these properties in the derivation above.218 Unity from Duality: Gravity.10) gives an indication of how T -duality works. called S-duality. a D3 is left invariant and a D5 is interchanged with an NS5. T -duality is a symmetry which has aspects peculiar to string theory [98].10) n/R denotes the quantized momentum of the center of mass of the string. 18. A “vertex” in which a D4 brane ends on a D6 brane. This non-perturbative symmetry. a closed bosonic string theory with one compact dimension whose radius in string units is R. The second term m2 R2 reﬂects the extended nature of the string.

which are part of the world-volume of the brane. 20). A T -duality involving a longitudinal dimension will transform a Dpbrane into a D(p − 1)-brane and will leave an NS5-brane intact. Berman and E. Rabinovici: Supersymmetric Gauge Theories 219 be generalized to an inﬁnite discrete symmetry and can be shown to actually be a gauge symmetry in the bosonic case. D and N denote Dirichlet and Neumann boundary conditions in the compact directions. In particular the transformation R → 1/R maps a type IIA string theory background with radius R to a type IIB background with radius 1/R and vice versa. respectively. those dimensions which are not part of the world-volume of the brane. T -duality acting on Dp branes. 19. In the presence of D-branes one naturally distinguishes between two types of compact dimensions: “longitudinal” dimensions. For supersymmetric string theories T -duality has some diﬀerent manifestations. . 19. This indicates that it persists non-perturbatively. Its eﬀect on a NS5-brane is more complicated and we will not need it in this lecture (Figs. and “transverse” dimensions. T -duality involving a transverse direction transforms a Dp-brane into a D(p+1)-brane. T Dp Rj ←− Dp T −→ E Rj ←− T T E T Ri D(p-1) Ri ‘Longitudinal Duality’ N −→ D E ‘Transverse Duality’ D(p+1) D −→ N E Fig.D.S.

and T -duality. x2 and x3 . Due to the odd number of T -duality transformations. The steps used in the proof are summarized in Figure 21. By performing a T -duality along three directions transverse to both the D3 and the D1branes. The construction sketched in this proof shows that any Dp-brane can end on any D(p + 2)-brane. T -duality acting on a NS5 brane. . The proof thus rests on the validity of both S. we obtain a D4-brane ending on a D6-brane. In a somewhat similar manner one can show that a D4-brane can end on a NS5-brane (Fig. There are many indications that the former is correct. The detailed action of the transverse duality is not indicated. Equipped with this information we can continue the proof of the existence of a D4 conﬁguration ending on a D6-brane. for example the directions x1 . 20.220 Unity from Duality: Gravity. Gauge Theory and Strings NS5 T T E NS5 NS5 T −→ E complicated ¡ ¡ ¡ ¡ e e e Fig. one passes from a IIB background to a type IIA background. one performs T -duality along two transverse direction. Starting from the by now established conﬁguration of a D1-brane ending on a D3-brane in type IIB string theory. x1 and x2 to obtain a D3-brane ending on a D5-brane. 22). and there is ﬁrmer evidence of the validity of T -duality.

Rabinovici: Supersymmetric Gauge Theories 221 F1 D3 −→ S D1 D3 −→ D6 06 T(1. 24). where + and − denote respectively directions longitudinal and . Berman and E. establishing that a D4 brane can indeed end on a D6 brane.D. A D4 brane ends on a NS5 brane.and T -duality transformations (called U -duality) is shown in Figure 23. leads to the desired conﬁguration in which a D4-brane ends on a NS5-brane. 21. D4 NS Fig. The spatial extension of the various branes is summarized in the table below. which is transverse to the newly formed D3brane but longitudinal to the NS5-brane. This series of S.2. 22. Let us consider ﬁrst the eﬀective low energy ﬁeld theory related to the conﬁguration in which the D4-brane is suspended between a NS5-brane and a D6-brane (Fig. An S-duality transformation brings us to a D3-brane ending on a NS5brane.3) D4 01236 0789 IIB IIB 0123789 IIA Fig. A T -duality along x3 . A combination of S and T duality transformations.S.

Gauge Theory and Strings D1 D3 06 −→ T(1. 24. D4 NS5 D6 Fig.222 Unity from Duality: Gravity. S.2) D5 D3 −→ 0612 S 0345 IIB 034512 IIB 0126 D3 −→ T(3) D4 0612 NS5 012345 IIB NS5 012345 IIA Fig. A D4 brane suspended between and NS5 and D6 branes. at best. The eﬀective ﬁeld theory contains no massless particles and is thus. some topological theory.and T -duality transformations establish that a D4 brane may end on a NS5 brane. . 23.

At both ends. x6 . (7. suspended between the D6 and the NS5 it allows no massless degrees of freedom to propagate on it.D. The eﬀective ﬁeld theory is N = 2 SUSY U (1) gauge theory. x8 and x9 . It is at best a topological ﬁeld theory. Although the D4-brane would have allowed the propagation of 5-dimensional photons had it been left on its own. x8 and x9 . x7 . On the D6 side. 25). Fig. A D4 brane suspended between two NS5 branes. x5 and x6 directions.S. D4 D6 N S5 x0123 + + + x4 − − + x5 − − + x6 = − − x7 − + − x8 − + − x9 − + −. The D4-brane left on its own could be displaced along the directions x4 . the D4-branes is locked in the x6 . We now suspend it between two NS5-branes which are extended in the same directions as the NS5 of the previous example [95] (Fig. Massless degrees of freedom can also be identiﬁed in a geometrical manner. We thus turn to another attempt to build an eﬀective four dimensional ﬁeld theory along the suspended D4-brane. x8 and x9 directions. Berman and E. Rabinovici: Supersymmetric Gauge Theories 223 transverse to the relevant brane. the D4-brane is frozen. x7 .11) The eﬀective ﬁeld theory should contain the massless degrees of freedom of the system. . On the NS5 side the D4 is locked along x6 . Therefore now it can be displaced in the x4 and x5 directions. Each possibility to displace the D4-brane along the D6 and the NS5 branes maintaining the shape of the conﬁguration corresponds to a massless particle. All in all. x7 . as shown in Figure 26. x5 . It cannot be displaced in a parallel fashion and therefore the eﬀective theory contains no massless particles whatsoever. 25. and = indicates that the brane is of ﬁnite extent in that direction. the D4-brane is locked in the x4 .

Gauge Theory and Strings Fig. N = 2 supersymmetry implies the existence of massless spin. (7. the gauge symmetry can be enhanced to U (NC ) by suspending NC D4-branes between the NS5-branes1 (Fig. it will be useful to study the eﬀective ﬁeld theory on a D4-brane suspended between two D6-branes (Fig. N = 2. N = 1. This corresponds to two massless spin-0 particles appearing in the eﬀective ﬁeld 1 theory. The eﬀective four dimensional gauge coupling constant is related to the eﬀective ﬁve dimensional gauge coupling constant in the usual Kaluza-Klein manner: 2 gYM. 26. U (1) supersymmetric gauge theory. 1 The actual symmetry turns out to be SU (NC ).4 = 2 gYM. .224 Unity from Duality: Gravity.3)) and (Eq.4 . The two scalar particle identiﬁed geometrically form part of the N = 2 vector multiplet. x9 directions will lead to the desired D = 4. An analysis using (Eqs. 99].12) 2 Changing the value of the separation ∆x6 amounts to rescaling gYM. Before pursuing this.2).2 and spin-1 particles as well. Rotating one of the NS5-branes from the x4 .4)) shows that in this conﬁguration one half of the supersymmetry generators of the single brane conﬁguration survive. Thus the conﬁguration above describes an eﬀective D = 4. That is. 28). spin 1/2 and spin 0 particles. That multiplet consists of spin 1. 27). (7. implementing an N = 2 supersymmetry in the eﬀective four dimensional theory. Thus the eﬀective ﬁeld theory contains at least two massless spin 0 particles. U (NC ) eﬀective gauge theory. The D4 brane can be parallely displaced along the x4 and x5 directions. As in the single brane case. 8 supercharges survive as symmetries. x5 directions to the x8 .5 · ∆x6 (7. This is discussed in [93. (7.

later we will show that . x8 and x9 . that is the low energy eﬀective theory is not an unbroken gauge theory. (7. 28.13) This conﬁguration also has 8 surviving supercharges and thus the eﬀective ﬁeld theory has N = 2 supersymmetry in D = 4. The eﬀective ﬁeld theory is a D = 4. It contains no massless gauge particles. These correspond to the allowed translations in the directions x7 . N = 2 U (NC ) SUSY gauge theory. At the case at hand. D4 D6 D6 Fig. Actually. Rabinovici: Supersymmetric Gauge Theories 225 D4 }NC NS5 NS5 Fig. It is important to note that the system contains no massless spin 1 degrees of freedom. the multiplet is actually the N = 2 hypermultiplet which contains four spin 0 and four spin 1/2 degrees of freedom. The eﬀective ﬁeld theory on the D4 brane is a D = 4. The spatial extension of the branes is: D4 D6 x0123 + + x4 − − x5 − − x6 = − x7 − + x8 − + x9 − +. N = 2 SUSY ﬁeld theory. Berman and E.D. An N = 2 supersymmetric multiplet would require either two or four spin 0 particles. 27. containg an N = 2 matter hypermultiplet. Similar considerations to the ones used above show that the eﬀective theory contains at least three massless spin 0 particles.S.

a rotation by diﬀerent angles would have given rise in ﬁeld theoretical language to a ﬁnite mass term. The eﬀective four dimensional theory is a U (1) gauge symmetry and has no moduli. x5 to the x8 . This is done by distributing NF D6-branes along the x6 direction (Fig. which denotes the separation of two D4-branes in the x4 . . as can be seen from the by now familiar geometrical considerations.5 An eﬀective D = 4. 30). The fact that the geometrical considerations were not suﬃcient to identify all four spin 0 particles shows us one of the limitations of the simple geometrical analysis. The gauge symmetry can be enhanced to U (NC ) by suspending NC D4-branes between the two diﬀerent NS5-branes (Fig. The mass of the W particles is again proportional to ∆D4(x4 . 7. one ﬁnds parameters in the brane picture with no clearly known ﬁeld theoretical interpretation and vice versa. N = 1 supersymmetric U (1) gauge theory. The complex number of moduli is immediately read out of the geometrical picture to be NC . x5 ). one notes that a separation of the D4-branes along the directions x4 . This is similar to what was described before in the case of the separations of NC inﬁnitely extended parallel Dp-branes. U(NC ) gauge theory with matter We construct now the conﬁguration leading to N = 1 supersymmetry. This agrees with a description by an eﬀective D = 4. Gauge Theory and Strings this conﬁguration will be part of the description of the Higgs phase of the supersymmetric gauge theory. The rotation of one of the NS5 from x4 . The ﬁnal ingredient needed is to add some ﬂavor to the eﬀective ﬁeld theory. x5 directions. namely A6 . 32). In Figure 29 we summarize the complex number of massless spin 0 particles appearing in any of the four conﬁgurations discussed until now. The fourth spin 0 particle can be identiﬁed in this case with the component of a compactiﬁed gauge ﬁeld. x9 directions corresponds to adding an inﬁnite mass term to the scalar ﬁelds in the adjoint representation. 31). N = 1.226 Unity from Duality: Gravity. The rotation leads to a conﬁguration with 4 surviving supercharges (Fig. At this stage of understanding of the gauge theory–brane correspondence. Returning to the N = 2 U (NC ) gauge conﬁguration. This coincides with algebraic analysis determining the complex number of massless spin 0 particles surviving the breaking of the gauge group. x5 leads to all rank preserving possible breakings of the gauge symmetry.

Berman and E.S. The spatial extension of the various branes is: x0123 + + + + x4 − − + − x5 − − + − x6 = − − − x7 − + − − x8 − + − + x9 − + − +.14) . A summary of the particle content of the low-energy eﬀective theories described in Section 7.D. D4 D6 N S5 N S5 (7.4. Rabinovici: Supersymmetric Gauge Theories 227 D4 D6 2 D6 N=2 Hyper-multiplet D4 NS5 1 NS5 Vector SUSY U(1) D4 NS5 0 D6 ‘Topological’ Fig. 29.

x5 directions. Gauge Theory and Strings ££ £ D4 £ £ NS5 45 £ £ £ £ £ £ NS5’ 89 Fig. The conﬁguration leading to a D = 4. £ £ NS5 45 £ £ £ £ £ £ ££ £ D4 }NC NS5’ 89 Fig. 30. and touching the D4-branes endows them with color. Their directionality is responsible for ¯ the appearance of both the NF and NF . N = 1 U (1) SUSY gauge theory. The matter ﬁelds appear as representations of diagonal vectorial subgroup of the ﬂavor group SU (NF ) × SU (NF ) × U (1). The way to obtain the Seiberg dual conﬁguration is essentially to move the position of the NS5-brane residing on the left hand part of the . This is as close to our goal as we will reach in this lecture. The conﬁguration leading to a D = 4. The masses of the squarks have a geometrical interpretation: they are proportional to the distance between the D4 and D6-branes along the x4 .228 Unity from Duality: Gravity. In [93] the reader will ﬁnd references for attempts to construct chiral conﬁgurations and to identify the full ﬂavor group. The NS5 extending also in the directions x8 . 31. Touching the D6-branes endows the open strings with ﬂavor. x9 is labeled NS5’. N = 1 U (NC ) SUSY gauge theory. The eﬀective ﬁeld theory is thus a four dimensional N = 1 U (NC ) supersymmetric gauge theory which has matter in the fundamental representation of the vector subgroup of the ﬂavor symmetry. Geometrically they are associated with strings connecting the D6-branes with the D4-branes.

32. 33. Berman and E. 28). 33).S. The conﬁguration shown has NC = 5 − 3 = 2. This is the dual conﬁguration to that of a N = 1 SUSY gauge theory with NC = 3 and NF = 5 (Fig. The NF ﬁelds are essentially the color-singlet particles appearing in the eﬀective ﬁeld theory corresponding to the brane conﬁguration (Fig. NF color singlet massless particles. The resulting new conﬁg¯ uration will have SU (NF − NC ) gauge symmetry and NF and NF colored 2 matter as well as NF . Adding D6 branes allows the existence of matter in the eﬀective ﬁeld theory. 7.D. Rabinovici: Supersymmetric Gauge Theories ££ £ s d d d d 229 £ £ £ £ £ £ £ £ NC D4 NS5 D6 D6 NS5’ Fig. 39). . in addition. E F) ¯ (NF .6 More pieces of information In order to be able to perform this displacement of the NS5-brane. which is exactly the 2 result of Seiberg (Fig. NC ) £ £ £ £ £ £ £ £ £ £ £ £ NS5’ D6 D6 D6 D6 D6 ' NF ˜ NC = 2 NS5 Fig. ﬂavor adjoint color singlets. Open strings ending on D4 and D6 branes carry both color and ﬂavor. one needs three more pieces of “information” as to the behavior of branes. conﬁguration all the way to the right hand part. 2 NF = 5 and. N £ £ £ £ £ £ c c i £ q £ £ £ £ §£¤ £ £ £ £ §£¤ §£¤ £ NF = 5 £ £ §£ ¤ §£ ¤ §£ ¤ £ £ §¤ §¤ §¤ §¤ £ £ £ £ £ £ £ £ £ s d £ d ˜ £ £ £ £ £ £ (NF .

34.D6 (x4 . a change in the relative displacement in the x4 . x5 ) mQ(Q) ˜ T x4 .D4 (x4 . x5 ) and ∆D4.D4 (x4 . The values of the masses of the W and squark particles is encoded in this ﬁgure. . A change in the relative position in the D4 and D6-branes in the x4 . x5 ) mWi NS5 D6 × s d d D4 d d D4 NS5 ∆D4. Both distances are indicated in the ﬁgure.230 Unity from Duality: Gravity. D4 T D4 c ∆D4. x5 ).D6 (x4 . They are proportional to the distances ∆D4. 34). Varying these distances smoothly changes the parameters of the ﬁeld theory. x5 E x6 Fig. x5 directions led to a change in the masses of the squarks (it is actually equivalent to a smooth change in the expectation value of the scalar coupled to fermions by a Yukawa coupling) (Fig. For example. respectively. x5 directions of the D4-branes in the N = 4 and N = 2 supersymmetric conﬁgurations led to a smooth change of the mass of the W ’s (or equivalently to smooth changes in the Higgs expectation values). Gauge Theory and Strings 1) “Motions of branes” Changes of the positions of branes correspond in some cases to smooth changes in ﬁeld theoretical parameters.

Consider an N = 2 . the two branes have to intersect. The N = 1 case has no Coulomb phase. Also the Higgs phase has a simple geometrical realization in terms of branes. the relative separation in the x6 direction. A change in the relative position along the x7 . 36). 35. x8 . Note that there is no particle content to the eﬀective theory on that D4-brane and thus no new degrees of freedom are created as the D4-brane is formed. x8 . Consider a conﬁguration consisting only of a NS5-brane and a D6-brane separated in the x6 direction (Fig. x8 and x9 directions of the positions of two NS5-branes corresponds to adding a FayetIliopoulos term in an N = 2 ﬁeld theoretical interpretation (Fig. Each such separation corresponds to appropriate components of the Fayet-Iliopoulos term in N = 2 SUSY U (NC ) gauge theory. However once they do. Rabinovici: Supersymmetric Gauge Theories 231 A change in the relative distance between the NS5-branes corresponded to a change in the coupling of the gauge theory on the D4-branes suspended between them. 37). 35). has no clear impact on the ﬁeld theory description). Eventually. What actually can be shown to occur [93. Berman and E. The same is true the N = 2 case.D. x9 ) Fig. The NS5 branes can be separated in the x7 . A similar interpretation can be given for the separation along the x7 direction between the NS5 and NS5 in the N = 1 case supersymmetric case. x8 . 95] (Fig. x9 directions. 95] is that as the NS5-brane “crosses” the D6-brane a D4-brane is created and suspended between the NS5 and D6-brane. 3) The brane realization of the Higgs phase We have seen that the Coulomb phase of N = 4 gauge theories is spanned by the separation among the D4-brane. it is clear that something more singular may result. Displacing the NS5brane for example in the x6 direction is a smooth motion as long as the two branes do not intersect (this is an example where a brane parameter.S. x9 T NS5 × © × NS5 E x6 ∆N S5 (x7 . In other cases changes in the position of branes could have more abrupt consequences. x7 . 2) Supersymmetry restrictions Supersymmetry allows that only a single D4-brane can be suspended between a NS5-brane and a D6-brane [93.

one starts the journey by setting the D = 4 N = 1 U (NC ) supersymmetric gauge theory with NF ﬂavors in the Higgs phase. one is able not only to reproduce the particle content required by Seiberg’s duality. To enforce that.232 Unity from Duality: Gravity. but also his analysis of comparing the dimensions of the moduli space at both ends of the duality transformation. from left to right in this ﬁgure. One can show that this breaking indeed represents the Higgs phenomenon. For the N = 1 case one needs also to realize that the eﬀective theory on a D4-brane suspended between a D6-brane and a NS5 -brane contains one massless complex spin 0 ﬁeld. In order to be able to trust the pictures as one moves the branes around it is advisable to be in a weak coupling situation as long as possible. A D4 brane is created when the NS5 brane crosses the D6 brane. 36. as well as comparing the numbers of the relevant operators. x5 directions being proportional to the expectation value of the scalar ﬁeld responsible for the Higgsing. . conﬁguration in which in addition of having a D4-brane suspended between two NS5-branes a D6-brane is inserted between the NS5-branes (Fig. Retaining the Higgs phase. This result is true also for the N = 1 conﬁgurations. x5 directions maintains the supersymmetry. the separation among the two D4-branes along the x4 . 38). The D4 brane carries no massless degrees of freedom. However the eﬀective ﬁeld theory on each of the D4-branes contains no spin 1 massless particles. The brane geometrical picture has a classical weakly coupled taste. The separation of the D4-branes on the D6-brane along the x4 . one massless spin 1/2 ﬁeld and no massless spin 1 ﬁelds. Gauge Theory and Strings £ £ NS5 45 £ £ £ £ £ £ £ £ E ' D6 789 D4 E£ £ £ £ £ £ £ £ £ £ ∆x6 D6 NS5 Fig. One is also able to show in the brane picture as Seiberg has done in ﬁeld theory that masses as one end of the duality pair correspond to expectation values on the other side.

S. 7. 2 An attempt is made to digitalize the movie. Berman and E. However.D. 93]). . SUSY allows only the existence of the conﬁguration a). The starting point is a conﬁguration depicted in Figure 39.7 Obtaining the dual ﬁeld theory At this stage in the lecture a movie2 composed out of the comic strips was shown. The reader is actually equipped by now to embark on this journey on her/his own (or by reading [68. I will brieﬂy mention the diﬀerent stages of the journey. Rabinovici: Supersymmetric Gauge Theories £ ££ 233 D4 £ £ £ £ £ £ £ NS5 a) £ £ D6 ££ D4 £ £ £ £ £ £ £ D4 £ NS5 b) £ £ D6 Fig. conﬁguration b) is allowed. As a navigational aid. 69. 37. Conﬁguration b) is disallowed. First stage: the NS5-brane is displaced across the D6-branes respecting the rule mentioned in 1) above concerning D4-brane creation. which has NF = 5 and NC = 3. This enabled an easy visualization of the continuous features of the duality transformation. if the NS5 is exchanged with an NS5’.

In the presence of additional D6 branes. the eﬀective theory would 1 contain massless spin-0 and spin. as appropriate for a Higgs phase (Fig. The Higgs phase of ﬁeld theory is realized by the D4 branes connecting the two NS5 branes. corresponds in some sense [68. as stated in 1) above. 28). This separation. the branes are separated along the x7 direction before they are made to coincide in the x6 direction. In order to avoid that. Second stage: the system is driven into the Higgs phase according to rule 3) above. Third stage: it is realized that if the NS5-brane will directly collide with the NS5 brane the gauge coupling will diverge. but no massless spin-1 degrees of freedom. 93] to the turning on of a Fayet-Iliopoulos coupling in ﬁeld theory. leaving the protected weak coupling regime. .2 degrees of freedom. breaking and separating on the D6 brane. The eﬀective ﬁeld theory in this example carries no massless degrees of freedom at all.234 Unity from Duality: Gravity. 38. At this stage. Gauge Theory and Strings NF = 1 D4 NC = 1 NS5 D6 NS5 D4 T D4 c Q NS5 D6 NS5 Fig. This indicated distance of separation corresponds to the expectation value of the Higgs matter ﬁeld. the appropriate reconnections of the branes shows that the conﬁguration represents the Higgs phase of a U (NF − NC ) gauge theory. 69.

S. Rabinovici: Supersymmetric Gauge Theories ' £ ££ D6 NC = 3 NF = 5 E 235 £ £ £ £ £ £ £ £ £ NS5 D6 D6 D6 D6 D6 NS5’ Fig. That does not really demonstrate that the two are identical in the infra-red or at all. The result here is simply the outcome of the NS5 moving according to the brane rules from the left of the NS5’ to its right. Other groups It can be shown that Sp(N ).D. the theory has 52 singlet complex spin-0 and spin. Taking a bus ride from Chamonix to Les Houches does not constitute a proof that the two cities are the same.1 massless particles. The ﬁnal outcome is shown = 5 ﬂavors and in Figure 33. 39. NF = 5 conﬁguration. A D = 4. In addition. 69. 93]. This is the realization of the dual the2 ory obtained by Seiberg. O(N ) and product gauge groups can be constructed as well [68. The simple procedure allows one to obtain the infra-red dual of many other brane conﬁgurations. N = 1 SUSY. 93]. Fourth stage: rejoining the branes in the x7 direction requires some quantum adjustments which we will indicate below. 7. . What has emerged naturally is 2 one gauge group out of another as well as the presence of NF color singlets. NC = 3. A more detailed analysis of the resulting massless particle spectrum (chiral ring) and its properties is needed to conclude infra-red duality [68.8 Concluding remarks Infrared duality What has actually been shown is that in a continuous motion starting from one N = 1 supersymmetric gauge theory one reaches another. the gauge theory has NF NC = 5 − 3 = 2 colors. Berman and E. 69.

and due to the compact nature of the x3 coordinate. It is more convenient to perform a longitudinal T -duality along the x3 direction. 99]. that is. such a vacuum state exists for any eﬀective supersymmetric gauge theory with no ﬂavors in more than three dimensions. For any ﬁnite radius. However the same system in D = 3 has no ground state [101]. the three dimensional gauge theory has no vacuum. say x3 .236 Unity from Duality: Gravity. This is indicated in the brane picture in the following manner: compactify one dimension. Quantum corrections The classical brane picture needs to be amended by quantum considerations [95. 100]. the type IIB eﬀective theory corresponds to a ﬁeld theory with more than three dimensions in type IIA. At this stage. 52]. For example. Only an inﬁnite type IIB radius will correspond to an eﬀective three dimensional gauge theory. these are additional postulates they have however immediate consequences in allowing a uniﬁed description of D = 4 and D = 3 gauge theories with 4 supercharges. Such a system has a ground state [15. M-theory context The brane–ﬁeld theory correspondence obtains even more geometrical features once embedded in M-theory [93. Once the IIB radius is inﬁnite. For example. Thus a four dimensional IIA system corresponds to a type IIB system with vanishing radius. recall that two D4-branes were not allowed to be suspended between an NS5 and a D6-brane. Gauge Theory and Strings Generalized infra-red dualities By allowing conﬁgurations with k NS5-branes and k NS5 -branes one can test new and old generalizations of Seiberg’s duality in the presence of a richer matter content and various Landau-Ginzburg-like interactions [68. 99. of the world-volume of the D4-brane. One could ascribe this to a quantum repulsive force between two D4-branes on the same side of a NS5-brane in a non-BPS conﬁguration. Due to the repulsive force between the D4-branes. 0 ≤ x3 ≤ 2πR. consider the D = 4 N = 1 U (NC ) gauge theory with no ﬂavors. the equilibrium state exists no more. The agreement between the dimensions of the moduli spaces of the two dual models was obtained in the presence of a quantum attractive interaction between two D4-branes on the opposite sides of a NS5-brane. The resulting system is a type IIB string theory whose third direction has radius 1/R. . a stable conﬁguration will form for any ﬁnite value of x3 . 69. 93].

A. [3] J. Zumino. . Perhaps a similar course can be followed for theories whose symmetry includes general coordinate invariance in analogy to Figures 4. Golfand and E. Phys. 13 (1971) 323 [Pisma Zh. A prototype of a phase diagram for a gravitational theory. acknowledges S.S. Elitzur. B 49 (1974) 52. Wess and B. D. 1990).R. Schwimmer for collobartions and many very valuable discussions. [2] J. The qualitative understanding of the phase structure of such systems will proceed their complete quantitative analysis. [4] P. Likhtman. Lett. Bagger and J. [102]. 5 and 8. Giveon. 13 (1971) 452]. The references given in this article though detailed are not complete. Supersymmetry and Supergravity.B. A.D. Wess. 2nd edition (Princeton University Press. D. 40. Berman and E. References [1] Y. 425.C.P. Teor. E. As a prototype for a phase diagram of gravity we propose the following Figure 40. A. Rabinovici: Supersymmetric Gauge Theories 237 gs Black Holes SUGRA 1 v HOLOGRAPHIC BOUND STRINGS V E Fig. Forge. 1992). Singapore (Singapore: World Scientiﬁc. Eksp. Kutasov and A. JETP Lett. The classiﬁcation of possible phases of gauge theories were understood well before concepts like conﬁnement could have been analysed analytically in four dimensional continuum theories. West. Introduction To Supersymmetry And Supergravity. acknowledges the inspirational atmosphere created by the groups of the Hebrew University and the Weizmann institute. Fiz. 8 Final remarks In this set of lectures we have had a panoramic vista of the rich structure of supersymmetric gauge theories.

J. Langacker and M. Weinberg. Fubini and E. Elitzur. W. Rev. Witten. P. Claus.P. Seiberg. B 51 (1974) 461. Nucl. de Boer and H. Nanopoulos. S. Phys. Rocek and W. Intriligator and N. Oz. Aharony. Nucl.238 [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] Unity from Duality: Gravity. Phys. R. P. Mat. Rept.V. Derix. Myerson and J. Fiz. See for example.lns. Nucl. 81 (1998) 4553. Phys. Rev. B 129 (1977) 493. Figueroa-O’Farrill. 62 (1998) 341. Van Proeyen. Phys. Pearson and J. Horn. D 52 (1995) 6157 [arXiv:hep-th/9507013]. Shigemitsu. Maldacena.cornell. Nucl. Gubser. Witten. J. Kogut. Amaldi. D 44 (1991) 817. Nucl. L. Introduction to Supersymmetry http://www. Private communication. Ukawa. D 13 (1976) 974. Nucl. Schwimmer. D 19 (1979) 3698. For a review see. Banks. Guth. Black holes and quantum mechanics [arXiv:hep-th/9902007]. Kumar. A 34 (1976) 569. Kelley and D. Fayet and J. M. S. Rabinovici. Nucl. T. Phys. Rept. P. Banks.edu/ argyres/phys661/ P. Rept. Kallosh. Front. Phys. Phys. Phys. Vol. Seiberg. Pashnev. Nilles. Lett. Phys. Rev. M. E. Phys. N. M. The Quantum Theory Of Fields. L. Rev. Kutasov and N. de Alfaro. 2000). Lett. Suppl. 110 (1984) 1.I. Seiberg. Proc. Ferrara. B 358 (1991) 600.R.B. E.M. Phys. Grisaru. Ooguri and Y. Furlan. B 188 (1981) 513. S. M. Forge and E. Lett. Parga. 323 (2000) 183 [arXiv:hep-th/9905111]. Nucl. D 21 (1980) 1013. Akulov and A. B 106 (1976) 219. Lett. Phys. J. E. Phys. K. O. Gates. Phys.P. P.x. BUSSTEPP lectures on supersymmetry [arXiv:hep-th/0109172]. Iliopoulos. UK: Univ. 56 (1983) 862. B 260 (1991) 447. B 96 (1975) 331. Rabinovici.S. Ellis.M. P. For a review see. B 119 (1982) 165. Yankielowicz. Phys. Lett. Phys. Nucl. 419. Phys.. Rev. P. Phys. S. Suppl. Yankielowicz. Suppl. Math. Schwimmer. Phys. Phys. Siegel. Phys. Peskin and S. Phys. O’Raifeartaigh. B 245 (1984) 17. B 260 (1991) 131. R. D 32 (1985) 927. B 255 (1985) 383. Weinberg. Lett. V. P. Furstenau. B 58 (1975) 67. Gauge Theory and Strings S. B 318 (1993) 469. Nucl. D 19 (1979) 3715. Phys. Lett.J. Susskind. Fayet and J. Illiopoulos. Phys. S. Elitzur. N. Rev.T.H. Phys. 56 (1983) 344. Phys. 3: Supersymmetry (Cambridge. Nucl. Rabinovici and A. M. Townsend and A. H. V. H. A. Rev. Phys. Proc. Weinstein and S. Rev. R. U. Windey and A. Rev. B 202 (1982) 253. D. West. O.C. Fayet and S. BC 45 (1996) 1 [hep-th/9509066]. Phys. D. Fubini and G. .E. A. Aharony. E. Phys.B. T.K. Proc. Nuovo Cim. Theor. Fayet. Sonnenschein. Pr. Kallosh. Rev. R. 58 (1983) 1 [arXiv:hep-th/0108200]. J. 68 (1998) 261 [arXiv:hep-th/9706168]. Weinberg. 32 (1977) 249. Teor. J. Rabinovici and A. S. Phys. D 19 (1979) 1277. Phys. S. S. Phys. P. Luo. D 20 (1979) 2619. S.

R. Nucl. B 342 (1995) 152 [hep-ph/9410203].V. Phys. Phys. Rossi and G. T. . (1995) 174 [hep-th/9512140]. B 138 (1978) 1. France. B 435 (1995) 129 [hep-th/9411149]. A. A. Davis. B 336 (1994) 409 [hep-th/9407106]. Intriligator. Phys. [70] N. Nucl. JHEP 0010 (2000) 011 [arXiv:hep-th/0009009]. Elitzur. Fradkin and S. [48] N. Witten. Nucl. B 83 (1979) 395. Witten. Phys. Leigh and N. Intriligator. Phys. Phys. Abbott and E. [52] D.P.L. Phys. Phys. Seiberg. B. B 160 (1979) 349. B 61 (1973) 45. Phys. Elitzur. Kutasov. D 19 (1979) 3682. Seiberg and E. [51] T. Lett. K. A. Phys. Seiberg.4 PRINT-80-0083 (UTRECHT) Lecture given at Cargese Summer Inst. [57] S. 26 – Sep. M. Argyres and M. 8. Forge. [60] K. Nucl. Khoze and M.G. Phys. E. B 353 (1995) 79 [hep-th/9504080]. B 431 (1994) 484 [arXiv:hep-th/9408099]. A. Schwimmer. Ahrenshoop Symp. Elitzur. Intriligator and N. V. Seiberg. Aug. Intriligator and E. Veneziano and S.D. Olesen. Phys. Forge. Elitzur. B 241 (1984) 493. Elitzur. Lett. Seiberg. Phys. in C79-08-26.S. Seiberg. [56] S. Shenker. (Cargese. A. Intriligator and N. [44] E. [46] S. Bilal. 162 (1988) 169. B 125 (1983) 487. Yankielowicz. Douglas. Phys. Taylor. Farhi. A. Nucl. Lett. B 444 (1995) 125 [hep-th/9503179]. Nucl. Elitzur. Phys. Nucl. [66] G. [58] K. A. [65] A. Y. Lett. [43] J. Phys. The calculus of many instantons [arXiv:hep-th/0206063]. A. Phys. Aﬄeck. Rabinovici and A. [53] I. Giveon and E. Giveon. R. 23 (1976) 245. Seiberg. [47] G. Lett. Phys. Rabinovici. Rabinovici. [68] S. 430 (1994) 485 [arXiv:hep-th/9407087]. Phys. Shenker. E.F. Forge. Rabinovici. Kutasov. [63] T. Rept. Veneziano. Phys. Pioline and E. Rabinovici. B 189 (1981) 547. [64] L. K. [49] N. [62] S. Nucl. B 431 (1994) 551 [hep-th/9408155]. 1979). Phys. Giveon. B 448 (1995) 93 [hep-th/9505062]. B 256 (1985) 557. [55] S. Lett. Phys. [54] K. Nucl. Konishi. Nucl. G. Seiberg and S. M. N. Phys. Nucl. S. Lett. B 400 (1997) 269 [hep-th/9702014]. Mattis. Phys. [69] S. Nucl. D 50 (1994) 1092 [hep-th/9403198]. [45] H.H. Banks and E. B 218 (1983) 493. [67] N. Nucl. Giveon and E. G. Rabinovici. B 505 (1997) 202 [hep-th/9704104]. Intriligator. Rabinovici: Supersymmetric Gauge Theories 239 [42] For similar structures in a string theory context see. B 205 (1982) 1. A. D. Giveon and D. Nucl. Forge. Lett. Mandelstam. Dine and N. Rev. Rabinovici. B 459 (1996) 160 [hep-th/9509130]. Dine and N. B 86 (1979) 283. [61] K. Rept. [72] A. Meurice. Rev. B 426 (1994) 19.C. [59] K.B. Duality in N = 2 SUSY SU (2) Yang-Mills Theory: A pedagogical introduction to the work of Seiberg and Witten [arXiv:hep-th/9601007]. Nielsen and P. [50] A. Phys. Rabinovici. ’t Hooft. B 379 (1996) 121 [hep-th/9603051]. Dorey. Berman and E.C. Erratum-ibid. Seiberg and E. ’t Hooft. Elitzur. Witten. Nucl. Giveon and E. Phys. Casher. Hollowood. Cardy and E.J. Nucl. [71] P. A. Phys. Phys. Amati.

Turin. Strassler. Petrova and I. Leigh and M. Einhorn.J. N. B. [87] E. 1994). M. Duﬀ. Giveon. G. Mod.D. unpublished. Phys. S.C. Rev. Petkova. B 500 (1997) 3 [hep-th/9703166]. [100] P. Hanany and E. West. Phys. Amit and E. 61 (1989).B. Brane Dynamics and Gauge Theory [hep-th/9802067]. [99] E. Giveon and D. B 318 (1993) 469 [hep-th/9309335]. Mack. Sov. B 461 (1996) 71 [hep-th/9511154]. B 524 (1998) 141 [arXiv:hep-th/9712040]. Seiberg and E. Nucl. J. Lett. Rev. Lett. Harvey [hep-th/9603082]. February 24-26. 53 (1977) 155. Nucl. Argyres. 220. Seiberg and E. [92] J. Lambert and P.I. Khuri and J.A. [102] J. Bander. Phys. Witten. Witten. Seiberg. Rabinovici. Phys. Symmetry And Simplicity in Physics. Symposium on The Occasion of Sergio Fubini’s 65th Birthday. Ronen Plesser. J. Witten. A. Phys. Singapore: World Scientiﬁc. Phys. Lett. [83] D. Phys. and references therein. Weinberg. 1987). Seiberg. Green. N. O. [78] M. Polchinski. Phys. B 257 (1985) 371. Nucl. B 363 (1995) 173 c [hep-th/9508043]. Superstring Theory (Cambridge University Press. N. . 52 (1984) 1188. Bardeen. Strassler. Phys.B. TASI Lectures on D-Branes [hep-th/9611050].M. Phys. see: A. Sciuto.R.I. Howe. Schwarz and E. Rabinovici. Witten. Kogan and E. Mack.L. Rept. Aharony.T. [95] A. Porrati and E.C. Harmonic Analysis On The N-Dimensional Lorentz Group And Its Application To Conformal Quantum Field Theory (Berlin. 284. Nucl. Giveon and M. Rept. see: J. Rabinovici.J.G. Rabinovici. Barbon. 1977).B. B 544 (1999) 104 [arXiv:hep-th/9809033].X. Saering and W. Published in “Erice 1999. Phys. Kol and S. Gauge Theory and Strings [73] T. Argyres. N. B 492 (1997) 152 [hep-th/9611230]. [80] For a review. Lambert and P. Yankielowicz. [94] E. [90] N. Todorov. Rev. Zeldovich. [91] For example. [82] S.C. Rabinovici. Dobrev.J. Phys. Phys. Nucl. B 196 (1982) 189.G. Lu. [88] N. B 418 (1998) 85 [arXiv:hep-th/9710034]. G. Non-perturbative gauge dynamics and strings. Math. [98] A.A. Phys.S. Phys. Rabinovici. [86] W. West. D 36 (1987) 562. M. Banks and A.240 Unity from Duality: Gravity. Lett. Roˇek. Aharony. N. Rabinovici. Moshe and M. Phys. [89] P.D. [84] E.K. Kogan and E. Nucl. B 447 (1995) 95 [arXiv:hep-th/9503121]. Phys. Nucl. B. Zaks. Basics and highlights in fundamental physics”.R. K. JHEP 9904 (1999) 015 [arXiv:hep-th/9902058]. Witten.B.H. 244 (1994) 77 [hep-th/9401139]. 259 (1995) 213 [hep-th/9412184]. Italy. Goldberg and E. Abel. in Proceedings. B 435 (1995) 129 [hep-th/9411149]. [101] O. 1994 (Singapore. Hanany.F. Bardeen. Phys. B 461 (1996) 71. S. [85] W.A. M. Phys. Barbon. [77] R. Nucl. [79] For some recent string theory results see. M. [75] V. [96] M. V. R. 280. [93] A. [74] G.L. [81] Ya. Plesser. Alberico and S. Intriligator. B 499 (1997) 67 [hep-th/9703110]. Rabinovici. Nucl. Phys. Uspekhi 11 (1968) 381. Nucl. Commun. and references therein. Kutasov. B 256 (1985) 499. Nucl. Phys. Seiberg and M. Phys. [76] P. Nucl. [97] M. I. I. JHEP 0206 (2002) 039 [arXiv:hep-th/0205090].

Allahabad 211019.LECTURE 3 AN INTRODUCTION TO DUALITY SYMMETRIES IN STRING THEORY A. Jhusi. Chhatnag Road. India . SEN Harish-Chandra Research Institute.

. . . . . . 296 299 . . . 299 . . . .2 Duality of dualities . . 301 . . . . . .3 Massless solitons and tensionless strings . . . . . . . . . . . . . . . 299 . . . . . .1 Construction of a dual pair of theories with eight supercharges . 4. 5 Precision test of duality: Spectrum of BPS states 5. . 4. . . . . . . . . . . . . . . . . . 246 . . . . . . .5 Other examples . . . . . . 309 8 M-theory 312 8. . . . . . . 254 3 Notion of duality symmetries in string theory 255 3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 Type I–SO(32) heterotic duality in D = 10 . . . . . . 6. . . . .3 Fiberwise duality transformation . 2. . . . . . on . . 260 261 264 268 272 274 .2 Testing duality conjectures . . . . . . 304 7 Duality in theories with less than sixteen supersymmetry generators 305 7. . . . . . . . . . . . . . . . . 5.3 Compactiﬁcation . . . . . . . . . . . . .4 Recovering higher dimensional dualities from lower dimensional ones . . . . . . . . . . . . .2 Interactions . . . . . . . . . . . . . . .2 Test of duality conjectures involving theories with eight supercharges . . . . . . . . . . . 306 7. . . . . . . . . . . 276 moduli . . . . 280 . . . 251 . . . . . . . . . . . . . . . . . .2 Self-duality of heterotic string theory on T 6 . . . . . . . . . . . 258 4 Analysis of low energy eﬀective ﬁeld theory 4. . . . . . . . . . . . . . . .3 Duality between heterotic on T 4 and type IIA 4. . . . . 255 3. Z) self-duality of type IIB in D = 10 . . . . . . . 4. . . 6 Interrelation between diﬀerent duality conjectures 6. . 312 8. . . . . . . . . . .1 Duality symmetries: Deﬁnition and examples . . . . . . 315 . . . Z) duality in type IIB on S 1 and D-branes . . . . . . 286 . .1 Combining non-perturbative and T -dualities . . . K3 . . 6. . . . . . . . 5. . . . . . . . . . . . . . . . . . 2. . . . . . . . .1 SL(2. . . . . . . . . . . . . . . .2 Compactiﬁcation of M-theory . 6. . .1 The spectrum . . . . . . .1 M-theory in eleven dimensions .4 SL(2. . . . . . . . . . . . . .Contents 1 Introduction 2 A brief review of perturbative 2. . . . . . . . . . . . . . . .2 SL(2. . . 243 string theory 245 . . . . Z) S-duality in heterotic on T 6 and multi-monopole spaces . . .

see [18]. These symmetries not only relate apparently different string theories. The key to this development is the discovery of duality symmetries. our understanding of string theory has undergone a dramatic change. Sen Abstract In this review I discuss some basic aspects of non-perturbative string theory. The topics include test of duality symmetries based on the analysis of the low energy eﬀective action and the spectrum of BPS states. I hope that this review will serve the limited purpose of initiating a person with a knowledge of perturbative string theory into this area.AN INTRODUCTION TO DUALITY SYMMETRIES IN STRING THEORY A. (For an introduction to perturbative string theory. However. Springer-Verlag 2002 . and an introduction to M-theory. this is solely due to the varied degree of familiarity that I have with this vast subject. instead of surveying all the important developments in this subject I shall try to explain the basic ideas with the help of a few simple examples. I apologise for the inherent bias in the choice of examples and the topics. relationship between diﬀerent duality symmetries. There are also many other reviews in this subject where more references can be found [1–17]. I have also not attempted to give a complete list of references. which relate the strong and weak coupling limits of apparently diﬀerent string theories. A complete list of references may be obtained by looking at the citations to some of the original papers in spires. Instead I have only included those references whose results have been directly used or mentioned in this article. In this review I shall try to give an introduction to this exciting subject. but give us a way to compute certain strong coupling results in one string theory by mapping it to a weak coupling result in a dual string theory.) c EDP Sciences. 1 Introduction During the last few years.

Gauge Theory and Strings The review will be divided into seven main sections as described below. 1. 4. This will in no way constitute an introduction to this subject. 5. M-theory: in this section I discuss the emergence of a new theory in eleven dimensions − now known as M-theory − from the strong coupling limit of type IIA string theory. a few examples of duality conjectures in string theory. Precision test of duality based on the spectrum of BPS states: in this section I shall discuss how one can device precision tests of various duality conjectures based on the analysis of the spectrum of a certain class of supersymmetric states in string theory. Throughout this article I shall work in units where = 1 and c = 1. 7. . Duality in theories with <16 supersymmetries: the discussion in Sections 4–6 is focussed on string theories with at least 16 supersymmetry generators. In this section I consider theories with less number of supersymmetries. I shall also discuss what we mean by relating diﬀerent dualities and try to formulate the rules that must be followed during such a derivation. A brief review of perturbative string theory: in this section I shall very brieﬂy recollect some of the results of perturbative string theory which will be useful to us in the rest of this article. I also discuss how compactiﬁcation of M-theory gives rise to new theories that cannot be regarded as perturbative compactiﬁcation of a string theory. Notion of duality symmetry: in this section I shall describe the notion of duality symmetry in string theory. Interrelation between various dualities: in this section I shall try to relate the various duality conjectures introduced in the Sections 3–5 by “deriving” them from a basic set of duality conjectures. Analysis of the low energy eﬀective action: in this section I shall describe how one arrives at various duality conjectures by analyzing the low energy eﬀective action of string theory. which correspond to N = 2 supersymmetry in four dimensions. 6. and the general procedure for testing these duality conjectures.244 Unity from Duality: Gravity. 3. 2. Speciﬁcally we shall focus our attention on theories with eight supercharges. at best it will serve as a reminder to a reader who is already familiar with this subject.

As we shall describe later. which appear as point-like objects to the present day experimentalists. 1. 2. A B Fig. where α has the dimension of (length)2 . we √ need to choose α to be of the order of 10−33 cm. and its appeal lies in its theoretical consistency. . are actually diﬀerent vibrational modes of strings. (a) (b) Fig. and b) an open string. Sen: Duality Symmetries 2 A brief review of perturbative string theory 245 String theory is based on the simple idea that elementary particles. but in order to correctly reproduce the strength of this√ interaction. known as string tension. a) A closed string. the typical size of a string is of the order √ of α ∼ 10−33 cm − a distance that cannot be resolved by present day experiments. Propagation of a closed string.A. Thus there is no direct way of testing string theory. The energy per unit length of the string. is parametrized as (2πα )−1 . Since α is the only length parameter in the theory. this theory automatically contains gravitational interaction between elmentary particles.

has ﬁnite extent. Typically each string theory contains a set of massless states and an inﬁnite tower of massive states. This gives an inﬁnite tower of states. Consider propagation of a string from a spacetime conﬁguration A to a space-time conﬁguration B. as we shall describe. Diﬀerent states of the string are obtained by acting on the Fock vacuum by these oscillators. E8 × E8 heterotic and SO(32) heterotic string theories respectively.1 The spectrum There are ﬁve known fully consistent string theories in ten dimensions. The massive string states typically have mass of the order of (10−33 cm)−1 ∼ 1019 GeV and are far beyond the reach of the present day accelerators. Since the spatial direction of the world-sheet theory has ﬁnite extent. The leads to ﬁve (apparently) diﬀerent consistent string theories in (9 + 1) dimensional space-time. However unlike a conventional quantum ﬁeld theory where the spatial directions have inﬁnite extent. In order to get a fully consistent string theory we need to add some internal fermionic degrees of freedom to the string and generalize the notion of area by adding new terms involving these fermionic degrees of freedom. 2b). They are known as type IIA. Here we give a brief description of the degrees of freedom and the spectrum of massless states in each of these . The amplitude for the propagation of the string from the space-time position A to space-time position B is given by the weighted sum over all world-sheet bounded by the initial and the ﬁnal locations of the string. The ﬁelds in this (1 + 1) dimensional quantum ﬁeld theory and the boundary conditions on these ﬁelds vary in diﬀerent string theories. It represents a compact circle if the string is closed (Fig. each world-sheet ﬁeld can be regarded as a collection of inﬁnite number of harmonic oscillators labelled by the quantized momentum along this spatial direction.246 Unity from Duality: Gravity. The weight factor is given by e−S where S is the product of the string tension and the area of the world-sheet. 2. We shall now brieﬂy describe the spectrum and interaction in various string theories and their compactiﬁcations. During this motion the string sweeps out a two dimensional surface in space-time. Gauge Theory and Strings The basic principle behind constructing a quantum theory of relativistic string is quite simple. known as the string world-sheet (see Fig. type I. 2a) and a ﬁnite line interval if the string is open (Fig. Thus the interesting part of the theory is the one involving the massless states. 1). type IIB. It turns out that this procedure by itself does not give rise to a fully consistent string theory. This (1 + 1) dimensional ﬁeld theory is known as the world-sheet theory. the dynamics of a point particle is described by quantum mechanics. here the spatial direction. Generalizing this we see that the ﬁrst quantized description of a string will involve a (1 + 1) dimensional quantum ﬁeld theory. In the ﬁrst quantized formalism. which labels the coordinate on the string.

We shall give the description in the so called light-cone gauge which has the advantage that all states in the spectrum are physical states. • NS-R where we put anti-periodic boundary condition on the leftmoving fermions and periodic boundary condition on the rightmoving fermions. and represent the eight transverse coordinates of a string moving in a nine dimensional space. In all there are four classes of states which need to be included in the spectrum: • NS-NS where we put anti-periodic boundary conditions on both the left. The eight scalar ﬁelds satisfy periodic boundary condition as we go around the circle. hence the spatial component of the world-sheet is a circle.and the right-moving fermions. we keep only about (1/4)th of the states in each sector by keeping only those states in the spectrum which have in them only even number of left-moving fermions and even number of right-moving fermions. • R-R where we put anti-periodic boundary conditions on both the left. 1. Type II string theories: in this case the world-sheet theory is a free ﬁeld theory containing eight scalar ﬁelds and eight Majorana fermions. It turns out that in order to get a consistent string theory we need to include in our theory diﬀerent classes of string states. Finally. These diﬀer from each other in . but at the end two possibilities remain. eight of them having left-handed chirality and the other eight having right-handed chirality.and right-moving fermions respectively.A. The fermions have a choice of having periodic or anti-periodic boundary conditions. The procedure has some ambiguity since in each of the four sectors we have the choice of assigning to the ground state either even or odd fermion number. We shall refer to these as left. It is useful to regard the eight Majorana fermions as sixteen Majorana-Weyl fermions. Consistency of string theory rules out most of these possibilities.and the right-moving fermions. some of which have periodic and some of which have anti-periodic boundary condition on the fermions. These eight scalar ﬁelds are in fact common to all ﬁve string theories. • R-NS where we put periodic boundary condition on the leftmoving fermions and anti-periodic boundary condition on the right-moving fermions. Both the type II string theories contain only closed strings. This is known as the GSO projection [125]. It is customary to refer to periodic boundary condition as Ramond (R) boundary condition [123] and anti-periodic boundary condition as Neveu-Schwarz (NS) boundary condition [124]. Sen: Duality Symmetries 247 theories.

the massless states from the RR sector of type IIB string theory consist of a scalar. Both superalgebras consist of 32 supersymmetry generators. Since the two theories differ only in their R-sector.and the right-moving Ramond ground states are carried out in an identical manner. an anti-symmetric rank two tensor ﬁeld. . The RR sector massless states of type IIA string theory consist of a vector.and the right-moving fermion number to the left. In the second possibility the GSO projections in the left. the world-sheet degrees of freedom of the string can be regarded as a collection of inﬁnite number of harmonic oscillators. and a rank three anti-symmetric tensor. 1 Although from string theory we get the spectrum of states. is characterized by several quantum numbers. In one possibility. and a scalar ﬁeld known as the dilaton1 .and the right-moving sector diﬀer from each other. This gives type IIB string theory. Typically states from the Ramond sector are in the spinor representation of the SO(9. Since the product of two spinor representation gives us back a tensor representation. Gauge Theory and Strings the following way. The spectrum of both these theories are invariant under space-time supersymmetry transformations which transform fermionic states to bosonic states and vice versa. it is useful to organise the spectrum in terms of ﬁelds.1) Lorentz algebra. On the other hand. and a rank four anti-symmetric tensor gauge ﬁeld satisfying the constraint that its ﬁeld strength is self-dual. For the creation operator associated with each oscillator we deﬁne the level as the absolute value of the number of units of world-sheet momentum that it creates while acting on the vacuum. In other words the spectrum of massless ﬁelds in string theory is identical to that of a free ﬁeld theory with these ﬁelds.248 Unity from Duality: Gravity. This theory is known as type IIA string theory. It will be useful to list the massless bosonic states in these two string theories. and those from the NS-R and R-NS sectors are fermionic. As has already been pointed out before. They constitute a symmetric rank two tensor ﬁeld. Often it is convenient to organise the inﬁnite tower of states in string theory by their oscillator level deﬁned as follows. The supersymmetry algebra for type IIB theory is known as the chiral N = 2 superalgebra and that of type IIA theory is known as the non-chiral N = 2 superalgebra. a rank two anti-symmetric tensor ﬁeld. whereas those from the NS sector are in the tensor representation. (The Fock vacuum. the assignment of the left. in turn. the states from the NS-NS and the RR sectors are bosonic. The total oscillator level of a state is then the sum of the levels of all the oscillators that act on the Fock vacuum to create this state. the NS sector bosonic states are the same in the two theories.

eight right-moving MajoranaWeyl fermions and thirty two left-moving Majorana-Weyl fermions. 2. . or all of them have anti-periodic boundary condition. In each sector we also have a GSO projection that keeps only those states in the spectrum which contain even number of left-moving fermions. Thus for example the total contribution to L0 will be given by the sum of the right-moving oscillator level and the contribution to L0 from the Fock vacuum.1) Lorentz algebra. Thus bosonic states come from states with NS boundary condition on the right-moving fermions and fermionic states come from states with R boundary condition on the right-moving fermions. Sen: Duality Symmetries 249 which are the momenta conjugate to the zero modes of various ﬁelds − modes carrying zero world-sheet momentum. in this case the boundary condition on the left-moving fermions do not aﬀect the Lorentz transformation properties of the state. Finally. The massless bosonic states in this theory consist of a symmetric rank two ﬁeld.(right-) moving oscillator level as the contribution to the oscillator level from the left. unlike in the case of type II string theories. We have as before NS and R boundary conditions as well as GSO projection involving the right-moving fermions.A. However. There are two possible boundary conditions on the left-moving fermions which give rise to fully consistent string theories. • E8 × E8 heterotic string theory: in this case we divide the thirty two left-moving fermions into two groups of sixteen each and 2 We should distinguish between world-sheet momentum. and the momenta of the (9 + 1) dimensional theory. an anti-symmetric rank two ﬁeld. Also as in the case of type II string theories. Heterotic string theories: the world-sheet theory of the heterotic string theories consists of eight scalar ﬁelds. the NS sector states transform in the tensor representation and the R sector states transform in the spinor representation of the SO(9. ¯ L0 and L0 include contribution from the oscillators as well as from the Fock vacuum. if E and P denote respectively the world-sheet energy ¯ and momentum2 then we deﬁne L0 = (E + P )/2 and L0 = (E − P )/2. a scalar ﬁeld known as the dilaton and a set of 496 gauge ﬁelds ﬁlling up the adjoint representation of the gauge group SO(32).(right-) moving bosonic and fermionic ﬁelds. They are: • SO(32) heterotic string theory: in this case we have two possible boundary conditions on the left-moving fermions: either all of them have periodic boundary condition. The latter are the the momenta conjugate to the zero modes of various bosonic ﬁelds in the world-sheet theory.) We can also separately deﬁne left.

an anti-symmetric rank two ﬁeld. and also even number of leftmoving fermions from the second group. The massless bosonic states in this theory consist of a symmetric rank two ﬁeld. In constructing type I string theory we keep only those states in the spectrum which are invariant under this world-sheet parity transformation. The relevant superalgebra is known as the chiral N = 1 supersymmetry algebra. This transformation is known as the world-sheet parity transformation. giving us back the two heterotic string theories. Gauge Theory and Strings use four possible boundary conditions. Using the bose-fermi equivalence in (1 + 1) dimensions. with the following two crucial diﬀerence. 3) all the leftmoving fermions in group 1 have periodic boundary conditions and all the left-moving fermions in group 2 have anti-periodic boundary conditions. Type I string theory: the world-sheet theory of type I theory is identical to that of type IIB string theory. In order to get a consistent string theory the momenta conjugate to these bosons must take discrete values. It turns out that there are only two consistent ways of quantizing the momenta.250 Unity from Duality: Gravity. • Type IIB string theory has a symmetry that exchanges the leftand the right-moving sectors in the world-sheet theory. (This symmetry is not present in type IIA theory since the GSO projection in the two sectors are diﬀerent). we can reformulate both the heterotic string theories by replacing the thirty two left-moving fermions by sixteen left-moving bosons. The spectrum of states in both the heterotic string theories are invariant under a set of space-time supersymmetry transformations. 1) all the left-moving fermions have periodic boundary condition 2) all the left-moving fermions have anti-periodic boundary condition. The world-sheet degrees of freedom are identical to those in the closed string sector. In each sector we also have a GSO projection that keeps only those states in the spectrum which contain even number of left-moving fermions from the ﬁrst group. a scalar ﬁeld known as the dilaton and a set of 496 gauge ﬁelds ﬁlling up the adjoint representation of the gauge group E8 × E8 . and has sixteen real generators. 3. • In type I string theory we also include open string states in the spectrum. Specifying the theory requires . 4) all the left-moving fermions in group 1 have anti-periodic boundary conditions and all the left-moving fermions from group 2 have periodic boundary conditions.

This spectrum is also invariant under the chiral N = 1 supersymmetry algebra with sixteen real supersymmetry generators.A.2 Interactions So far we have discussed the spectrum of string theory. and appropriate boundary conditions on the fermions. we would like to know how to compute a scattering amplitude involving various string states. Fig. It turns out that there is a unique way of introducing interaction in string theory. 3. If we imagine the time axis running from left to right. In particular. − the analog of a tree diagram in . Consider for example a scattering involving four external strings. situated along some speciﬁc curves in space-time. We put Neumann boundary condition on the eight scalars. 2. One such surface is shown in Figure 3. but in order to fully describe the theory we must also describe the interaction between various particles in the spectrum. and 496 gauge ﬁelds in the adjoint representation of SO(32) from the open string sector. A string world-sheet bounded by four external strings. S being the string tension multiplied by the generalized area of this surface (taking into account the fermionic degrees of freedom of the world-sheet). The spectrum of massless bosonic states in this theory consists of a symmetric rank two tensor and a scalar dilaton from the closed string NS sector. Sen: Duality Symmetries 251 us to specify the boundary conditions on the various ﬁelds. then this diagram represents two strings joining into one string and then splitting into two strings. an anti-symmetric rank two tensor from the closed string RR sector. The prescription for computing the scattering amplitude is to compute the weighted sum over all possible string world-sheet bounded by the four strings with weight factor e−S .

This introduces an arbitrary parameter in string theory. which then splits into two and joins again. we should reproduce the S-matrix elements involving the massless states of string theory. By deﬁnition this eﬀective action is such that if we compute the tree level scattering amplitude using this action. In practice there is an extremely eﬃcient method of doing this computation using the so called vertex operators. In general such an action will have to . known as the string coupling constant. Gauge Theory and Strings ﬁeld theory. The most convenient way to summarize the result of this computation in any string theory is to specify the eﬀective action.252 Unity from Duality: Gravity. all of these scattering amplitudes in string theory are ultraviolet ﬁnite. once the relative normalization between these two diagrams is ﬁxed. Fig. The more relevant quantity is the scattering amplitude where the external strings are in the eigenstates of the energy and momenta operators conjugate to the coordinates of the (9 + 1) dimensional space-time. It turns out that unlike in quantum ﬁeld theory. Thus besides the dimensionful parameter α . This is done by simply taking the convolution of the above scattering amplitude with the wavefunctions of the strings corresponding to the external states. The relative normalization between the contributions from these two diagrams is not determined by any consistency requirement. Out main interest will be in the scattering involving the external massless states. A more complicated string world-sheet. the relative normalization between all other diagrams is ﬁxed due to various consistency requirement. both these parameters can be absorbed into deﬁnitions of various ﬁelds in the theory. What we have described so far is the computation of the scattering amplitude with ﬁxed locations of the external strings in space-time. 4. This represents two strings joining into one string. and ﬁnally splits into two strings. A more complicated surface is shown in Figure 4. However. This is one of the major achievements of string theory. As we shall see later. This is the analog of a one loop diagram in ﬁeld theory. string theory has a single dimensionless coupling constant.

On the other hand the actions for the three heterotic string theories correspond to another set of well-known supersymmetric theories in ten dimensions. The perturbative eﬀective action does not have any potential for Φ. (2. One expects that in a realistic string theory where supersymmetry is spontaneouly broken. (2. The emergence of gravity in all the ﬁve string theories is the most striking result in string theory. and hence Φ can take arbitrary value. shows that string theory gives us a ﬁnite quantum theory of gravity. the eﬀective action will have an invariance under the simultaneous rescaling of α and the metric gµν : α → λα . The actions for the type IIA and type IIB string theories correspond to those of two well known supergravity theories in ten space-time dimensions. g S → eC g S . gµν → λgµν . This.2) . Since α has the dimension of (length)2 and is the only dimensionful parameter in the theory. Using this scaling property. Here Φ denotes the dilaton ﬁeld. − so called because this gives the dominant contribution if we want to evaluate the scattering amplitude when all the external particles have small energy and momenta. gS can be absorbed in Φ. gS denotes the string coupling. the dimensionless coupling constant in string theory is related to the vacuum expectation value Φ of Φ. In a similar vain one can argue that in string theory even the string tension. or equivalently the parameter α . For type I and the SO(32) heterotic string theories the Yang-Mills gauge group is SO(32) whereas for the E8 × E8 heterotic string theory the gauge group is E8 × E8 . called type IIA and type IIB supergravity theories respectively. The low energy eﬀective action for all ﬁve string theories have been found. but we can organise these terms by examining the number of space-time derivatives that appear in a given term in the action. The eﬀective action of all ﬁve string theories are invariant under the transformation Φ → Φ − 2C.1) together with possible rescaling of other ﬁelds. and hence Φ will be determined uniquely. and C is an arbitrary constant.A. there will be a potential for Φ. We shall explicitly write down the low energy eﬀective action of some of the string theories in Section 4. has no physical signiﬁcance. Sen: Duality Symmetries 253 contain inﬁnite number of terms. Terms with the lowest number of derivatives constitute the low energy eﬀective action. Its origin can be traced to the existence of the symmetric rank two tensor state (the graviton) in all these theories. combined with the result on ﬁniteness of scattering amplitudes. − N = 1 supergravity coupled to N = 1 super Yang-Mills theory. Put another way.

At the end both approaches yield identical results. The idea is to take the (9 + 1) dimensional space-time as the product of a (9 − d) dimensional compact manifold M with euclidean signature and a (d + 1) dimensional Minkowski space Rd.1 . Instead of going via the eﬀective action. since this guarantees vanishing of the cosmological constant. 2. Consistency of string theory puts restriction on the kind of manifold on which the string can propagate. Gauge Theory and Strings together with possible rescaling of other ﬁelds. These states are known as winding states and play a crucial role in the analysis of duality symmetries. in the limit when the size of the compact manifold is suﬃciently small so that the present day experiments cannot resolve this distance. The simplest class of compact manifolds. Choosing d = 3 will give us a (3 + 1) dimensional theory. We shall discuss these two rescalings in detail in Section 4. and hence consistency of the corresponding string theory order by order in perturbation theory. Using this scaling symmetry α can be absorbed into the deﬁnition of gµν . instead of in ﬂat ten dimensional space-time. tori of diﬀerent dimensions. the world will eﬀectively appear to be (d + 1) dimensional. For such a states. K3. The eﬀect of this compactiﬁcation is to periodically identify some of the bosonic ﬁelds in the string world-sheet ﬁeld theory − the ﬁelds which represent coordinates tangential to the compact circles.3 Compactiﬁcation So far we have described ﬁve diﬀerent string theories.g. one can also directly describe these compactiﬁed theories as string theories. we also go once around the compact circle. Since our world is (3 + 1) dimensional.1 . as we go once around the string.1. it must satisfy the equations of motion of the eﬀective ﬁeld theory that comes out of string theory. but they all live in ten space-time dimensions. One eﬀect of this is that the momentum carried by any string state along any of these circles is quantized in units of 1/R where R is the radius of the circle. But that is another novel eﬀect: we now have new states that correspond to strings wrapped around a compact circle. This modiﬁes the world-sheet theory to an interacting non-linear σ-model instead of a free ﬁeld theory.254 Unity from Duality: Gravity. However one can construct string theories in lower dimensions using the idea of compactiﬁcation. One also normally considers only those manifolds which preserve part of the space-time supersymmetry of the original ten dimensional theory. Of course we cannot choose any arbitrary manifold M for this purpose. these are not realistic string theories. For this one needs to modify the string world-sheet action in such a way that it describes string propagation in the new manifold M × Rd. There are many known examples of manifolds satisfying these restrictions e. . on which we shall focus much of our attention in the rest of this article. Then. are tori − product of circles. Calabi-Yau manifolds etc.

We also saw that we can get many diﬀerent string theories in lower dimensions by compactifying these ﬁve theories on appropriate manifold M. • various other background ﬁelds. We begin by introducing the notion of duality in string theory. This situation has been illustrated in Figure 5. • Shape and size of M (information contained in the metric). This situation is illustrated in Figure 6.g. 5. Inside the moduli space of the theory there is a certain region where the string coupling is weak and perturbation theory is valid. • String coupling constant (related to the vacuum expectation value of the dilaton ﬁeld). Fig. whereas the white region denotes the strong coupling region. In general this equivalence relation maps the weak coupling region of one theory to the strong coupling region of the second theory and vice versa. and the way of avoiding these diﬃculties.A. Each of these theories is parametrized by a set of parameters known as moduli3 e. The shaded region denotes the weak coupling region. . the diﬃculties in testing them.1 Duality symmetries: Deﬁnition and examples As was described in the last section. Sen: Duality Symmetries 3 Notion of duality symmetries in string theory 255 In this section I shall elaborate the notion of duality symmetries. there are ﬁve consistent string theories in ten space-time dimensions. A schematic representation of the moduli space of a string theory. String duality provides us with an equivalence map between two diﬀerent string theories. Elsewhere the theory is strongly coupled. 3. 3 In string theory these moduli are related to vacuum expectation values of various dynamical ﬁelds and are expected to take deﬁnite values when supersymmetry is broken.

Thus for example. A schematic representation of the duality map between the moduli spaces of two diﬀerent string theories. Fig. K are two compact manifolds. • Type IIA string theory compactiﬁed on K3 and heterotic string theory compactiﬁed on a four dimensional torus T 4 are conjectured to be dual to each other [19. tree level results in one theory might include perturbative and non-perturbative corrections in the dual theory.1 denotes (10 − n) dimensional Minkowski space. let us give a few examples of dual pairs: • Type I and SO(32) heterotic string theories in D = 10 are conjectured to be dual to each other [19–22]. Duality relates weak and strong coupling regions of the same theory.256 Unity from Duality: Gravity. typically perturbation expansions get mixed up. 23–26]4 . 4 Throughout this article a string theory on M will mean string theory in the background M × R9−n. Under duality. Gauge Theory and Strings A on K B on K Fig. Under this duality the weak coupling region of the ﬁrst theory (denoted by the shaded region) gets mapped to the strong coupling region of the second theory and vice versa. A on K and B on K . Also under duality. and R9−n. 7. Before we proceed.1 where n is the real dimension of M. and K. where A and B are two of the ﬁve string theories in ten dimensions. many of the elementary string states in one theory get mapped to solitons and their bound states in the dual theory. Schematic representation of the moduli space of a self-dual theory. 6. .

For example. duality transformations form a symmetry group that acts on the moduli space of the theory. In each case the shaded region denotes weak coupling region as usual. A schematic representation of the moduli spaces of a chain of theories related by duality. For example. 9. there are special cases where the situation is a bit diﬀerent. Examples of T -duality relating a weakly coupled theory to a diﬀerent or the same weakly coupled theory. 8. type IIB . Sen: Duality Symmetries 257 Fig. Fig.A. we can have: • Self-duality: here duality gives an equivalence relation between diﬀerent regions of the moduli space of the same theory. Although duality in general relates the weak coupling limit of one theory to the strong coupling limit of another theory. In this case. as illustrated in Figure 7.

either of the two heterotic string theories compactiﬁed on a circle of radius R is dual to the same theory compactiﬁed on a circle of radius R−1 at the same value of the coupling constant. But in string theory we only know how to deﬁne the theory perturbatively at weak coupling. at least to all orders in perturbation theory. Gauge Theory and Strings string theory in D = 10 is conjectured to have an SL(2. type IIA string theory compactiﬁed on K3 is related to heterotic string theory compactiﬁed on T 4 . see [27]. All T -duality symmetries in string theory can be “proved” this way. and hence can be tested using string perturbation theory. Furthermore. On the other hand. it allows us to study a strongly coupled string theory by mapping it to a weakly coupled dual theory whenever such a dual theory exists. SO(32) heterotic string theory compactiﬁed on T 4 is related to type I string theory compactiﬁed on T 4 . Thus these three theories are related by a chain of duality transformations. Thus for example. Thus it would seem impossible to prove or test any duality conjecture in string theory5 . duality typically relates a weakly coupled string theory to a strongly coupled string theory. Also.258 Unity from Duality: Gravity. by relating various apparently unrelated (compactiﬁed) string theories. 3. as illustrated in Figure 9. due to the equivalence of the SO(32) heterotic and type I string theory in ten dimensions.) In a generic situation duality can relate not just two theories. Z) self-duality group [23]. but a whole chain of theories. From this discussion we see that the presence of duality in string theory has two important consequences. it reduces the degree of non-uniqueness of string theory. First of all. Thus in order to prove/test duality we must be able to analyze at least one of the theories at strong coupling. For example. This is where supersymmetry comes to our 5 Note that this problem is absent for T -duality transformations which relates two weakly coupled string theories.2 Testing duality conjectures Let us now turn to the question of testing duality. • T -duality: in this case duality transformation maps the weak coupling region of one theory to the weak coupling region of another theory or the same theory as illustrated in Figure 8. (For a review of this subject. As a result the duality map does not mix up the perturbation expansions in the two theories. . type IIA string theory compactiﬁed on a circle of radius R is dual to IIB string theory compactiﬁed on a circle of radius R−1 at the same value of the string coupling. As we have already emphasized.

A. Sen: Duality Symmetries

259

rescue. Supersymmetry gives rise to certain non-renormalization theorems in string theory, due to which some of the weak coupling calculations can be trusted even at strong coupling. Thus we can focus our attention on such “non-renormalized” quantities and ask if they are invariant under the proposed duality transformations. Testing duality invariance of these quantities provides us with various tests of various duality conjectures, and is in fact the basis of all duality conjectures. The precise content of these non-renormalization theorems depends on the number of supersymmetries present in the theory. The maximum number of supersymmetry generators that can be present in a string theory is 32. This gives N = 2 supersymmetry in ten dimensions, and N = 8 supersymmetry in four dimensions. Examples of such theories are types IIA or type IIB string theories compactiﬁed on n dimensional tori T n . The next interesting class of theories are those with 16 supersymmetry generators. This corresponds to N = 1 supersymmetry in ten dimensions and N = 4 supersymmetry in four dimensions. Examples of such theories are types IIA or type IIB string theories compactiﬁed on K3 × T n , heterotic string theory compactiﬁed on T n , etc. Another class of theories that we shall discuss are those with eight supersymmetry generators, e.g. heterotic string theory on K3 × T n , type IIA or IIB string theory on six dimensional Calabi-Yau manifolds, etc. For theories with 16 or more SUSY generators the non-renormalization theorems are particularly powerful. In particular, • Form of the low energy eﬀective action involving the massless states of the theory is completely ﬁxed by the requirement of supersymmetry (and the spectrum) [28]. Thus this eﬀective action cannot get renormalized by string loop corrections. As a result, any valid symmetry of the theory must be a symmetry of this eﬀective ﬁeld theory; • These theories contain special class of states which are invariant under part of the supersymmetry transformations. They are known as BPS states, named after Bogomol’nyi, Prasad and Sommerfeld. The mass of a BPS state is completely determined in terms of its charge as a consequence of the supersymmetry algebra. Since this relation is derived purely from an analysis of the supersymmetry algebra, it is not modiﬁed by quantum corrections. Furthermore it can be argued that the degeneracy of BPS states of a given charge does not change as we move in the moduli space even from weak to strong coupling region [29]. Thus the spectrum of BPS states can be calculated from weak coupling analysis and the result can be continued to the strong coupling region. Since any valid symmetry of the theory must be a symmetry of the spectrum of BPS states, we can use this to design non-trivial tests of duality [1].

260

Unity from Duality: Gravity, Gauge Theory and Strings

For theories with eight supersymmetries the non-renormalization theorems are less powerful. However, even in this case one can design non-trivial tests of various duality conjectures. We shall discuss these in Section 7. 4 Analysis of low energy eﬀective ﬁeld theory

In this section I shall discuss tests of various dualities in string theories with ≥16 supersymmetries based on the analysis of their low energy eﬀective action. As has been emphasized in the previous section, the form of this low energy eﬀective action is determined completely by the requirement of supersymmetry and the spectrum of massless states in the theory. Thus it does not receive any quantum corrections, and if a given duality transformation is to be a symmetry of a string theory, it must be a symmetry of the corresponding low energy eﬀective action. Actually, since the low energy eﬀective action is to be used only for deriving the equations of motion from this action, and/or computing the tree level S-matrix elements using this action, but not to perform a full-ﬂedged path integral, it is enough that only the equations of motion derived from this action are invariant under duality transformations. (This also guarantees that the tree level S-matrix elements computed from this eﬀective action are invariant under the duality transformations.) It is not necessary for the action itself to be invariant. Throughout this article we shall denote by Gµν the string metric − the metric that is used in computing the area of the string world-sheet embedded in space time for calculating string scattering amplitudes. For a string theory compactiﬁed on a (9 − d) dimensional manifold M, we shall denote by Φ the shifted dilaton, related to the dilaton Φ(10) of the ten dimensional string theory as Φ = Φ(10) − ln V , (4.1)

where (2π)9−d V is the volume of M measured in the ten dimensional string (10) metric. The dilaton is normalized in such a way that e Φ corresponds to the square of the closed string coupling constant in ten dimensions6 . gµν will denote the canonical Einstein metric which is related to the string metric by an appropriate conformal rescaling involving the dilaton ﬁeld, gµν = e− d−1 Φ Gµν .

2

(4.2)

We shall always use this metric to raise and lower indices. The signature of space-time will be taken as (−, +, · · · +). Finally, all ﬁelds will be made dimensionless by absorbing appropriate powers of α in them.

6Φ

is related to the more commonly normalized dilaton φ by a factor of two: Φ = 2φ.

A. Sen: Duality Symmetries

261

We shall now consider several examples. The discussion will closely follow references [1, 19, 23]. For a detailed review of the material covered in this section, see reference [15]. 4.1 Type I–SO(32 ) heterotic duality in D = 10 In SO(32) heterotic string theory, the massless bosonic states come from (H) the NS sector of the closed heterotic string, and contains the metric gµν , (H) the dilaton Φ(H) , the rank two anti-symmetric tensor ﬁeld Bµν , and gauge (H)a ﬁelds Aµ (1 ≤ a ≤ 496) in the adjoint representation of SO(32). The low energy dynamics involving these massless bosonic ﬁelds is described by the N = 1 supergravity coupled to SO(32) super Yang-Mills theory in ten dimensions [114]. The action is given by [102]: S (H) = 1

2 (2π)7 (αH )4 gH

1 d10 x −g (H) R(H) − g (H)µν ∂µ Φ(H) ∂ν Φ(H) 8

(H) 1 (H) (H) − g (H)µµ g (H)νν e−Φ /4 Tr(Fµν Fµ ν ) 4 (H) 1 (H) (H) − g (H)µµ g (H)νν g (H)ρρ e−Φ /2 Hµνρ Hµ ν ρ , 12

**(4.3) where R(H) is the Ricci scalar, Fµν denotes the non-abelian gauge ﬁeld strength,
**

(H) Fµν = ∂µ A(H) − ∂ν A(H) + ν µ (H)

2 [A(H) , A(H) ] , ν αH µ

(H)

(4.4)

Tr denotes trace in the vector representation of SO(32), and Hµνρ is the (H) ﬁeld strength associated with the Bµν ﬁeld:

(H) Hµνρ

1 1 (H) (H) = ∂µ Bνρ − Tr A(H) Fνρ − µ 2 3 +cyclic permutations of µ, ν, ρ .

2 (H) (H) (H) A [Aν , Aρ ] αH µ (4.5)

2παH and gH are respectively the inverse string tension and the coupling constant of the heterotic string theory. The rescalings (2.1), (2.2) take the following form acting on the complete set of ﬁelds: g H → eC g H ,

(H) (H) Bµν → Bµν ,

Φ(H) → Φ(H) − 2C, A(H)a → A(H)a , µ µ

(H) (H) gµν → eC/2 gµν

(4.6)

262

**Unity from Duality: Gravity, Gauge Theory and Strings αH → λαH ,
**

(H) (H) Bµν → λBµν ,

Φ(H) → Φ(H) ,

(H) (H) gµν → λgµν

A(H)a → λ1/2 A(H)a , µ µ

(4.7)

Since gH and αH can be changed by this rescaling, these parameters cannot have a universal signiﬁcance. In particular, we can absorb gH and αH into the various ﬁelds by setting e−C = gH and λ = (αH )−1 in (4.6), (4.7). This is equivalent to setting gH = 1 and αH = 1. In this notation the physical (H) coupling constant is given by the vacuum expectation value of eΦ /2 , and the ADM mass per unit length of an inﬁnitely long straight string, measured (H) (H) (H) in the metric e Φ /4 gµν that approaches the string metric Gµν far away (H) we can get all possible from the string, is equal to 1/2π. By changing Φ values of string coupling, and using a metric that diﬀers from the one used here by a constant multiplicative factor, we can get all possible values of the string tension. For αH = 1 and gH = 1 equations (4.3)–(4.5) take the form: S (H) = 1 1 d10 x −g (H) R(H) − g (H)µν ∂µ Φ(H) ∂ν Φ(H) 7 (2π) 8 1 (H)µµ (H)νν −Φ(H) /4 (H) (H) − g g e Tr(Fµν Fµ ν ) 4 (H) 1 (H) (H) − g (H)µµ g (H)νν g (H)ρρ e−Φ /2 Hµνρ Hµ ν ρ , 12 (4.8)

(H) Fµν = ∂µ A(H) − ∂ν A(H) + ν µ

√

2[A(H) , A(H) ] , µ ν

(4.9)

(H) Hµνρ

**√ 1 2 (H) (H) (H) (H) (H) A [Aν , Aρ ] = − Tr Aµ Fνρ − 2 3 µ +cyclic permutations of µ, ν, ρ .
**

(H) ∂µ Bνρ

(4.10)

Let us now turn to the type I string theory. The massless bosonic states in type I theory come from three diﬀerent sectors. The closed string Neveu(I) Schwarz − Neveu-Schwarz (NS) sector gives the metric gµν and the dilaton Φ(I) . The closed string Ramond-Ramond (RR) sector gives an anti(I) symmetric tensor ﬁeld Bµν . Besides these, there are bosonic ﬁelds coming from the NS sector of the open string. This sector gives rise to gauge ﬁelds (I)a Aµ (a = 1, . . . 496) in the adjoint representation of the group SO(32). (The superscript (I) refers to the fact that these are the ﬁelds in the type I string theory.) The low energy dynamics is again described by the N = 1 supergravity theory coupled to SO(32) super Yang-Mills theory [115]. But

A. Sen: Duality Symmetries

263

it is instructive to rewrite the eﬀective action in terms of the type I variables. For suitable choice of the string tension and the coupling constant, this is given by [102] S (I) = 1 1 d10 x −g (I) R(I) − g (I)µν ∂µ Φ(I) ∂ν Φ(I) 7 (2π) 8 1 (I)µµ (I)νν Φ(I) /4 (I) (I) − g g e Tr(Fµν Fµ ν ) 4 (I) 1 (I) (I) − g (I)µµ g (I)νν g (I)ρρ eΦ /2 Hµνρ Hµ ν ρ , 12 (4.11) where R(I) is the Ricci scalar, Fµν denotes the non-abelian gauge ﬁeld strength, √ (I) Fµν = ∂µ A(I) − ∂ν A(I) + 2[A(I) , A(I) ] , (4.12) ν µ µ ν and Hµνρ is the ﬁeld strength associated with the Bµν ﬁeld: √ 1 2 (I) (I) (I) (I) (I) (I) A [Aν , Aρ ] Hµνρ = ∂µ Bνρ − Tr A(I) Fνρ − µ 2 3 µ +cyclic permutations of µ, ν, ρ .

(I) (I) (I)

(4.13)

For both, the type I and the SO(32) heterotic string theory, the low energy eﬀective action is derived from the string tree level analysis. However, to this order in the derivatives, the form of the eﬀective action is determined completely by the requirement of supersymmetry for a given gauge group. Thus neither action can receive any quantum corrections. It is straightforward to see that the actions (4.8) and (4.11) are identical provided we make the identiﬁcation: Φ(H) = −Φ(I) ,

(H) (I) Bµν = Bµν , (H) (I) gµν = gµν

A(H)a = A(I)a . µ µ

(4.14)

This led to the hypothesis that the type I and the SO(32) heterotic string theories in ten dimensions are equivalent [19]. One can ﬁnd stronger evidence for this hypothesis by analysing the spectrum of supersymmetris states, but the equivalence of the two eﬀective actions was the reason for proposing this duality in the ﬁrst place. Note the − sign in the relation between Φ(H) and Φ(I) in equation (4.14). Recalling that e Φ /2 is the string coupling, we see that the strong coupling limit of one theory is related to the weak coupling limit of the other theory and vice versa.

264

Unity from Duality: Gravity, Gauge Theory and Strings

From now on I shall use the unit α = 1 for writing down the eﬀective action of all string theories. Physically this would mean that the ADM mass per unit length of a test string, measured in the metric e2 Φ /(d−1) gµν that agrees with the string metric Gµν deﬁned in (4.2) far away from the test string, is given by 1/2π. In future we shall refer to the ADM mass of a particle measured in this metric as the mass measured in the string metric. 4.2 Self-duality of heterotic string theory on T 6 In the previous subsection we have described the massless bosonic ﬁeld content of the ten dimensional SO(32) heterotic string theory. When we compactify it on a six dimensional torus, we can get many other massless scalar ﬁelds from the internal components of the metric, the anti-symmetric tensor ﬁeld and the gauge ﬁelds in the Cartan subalgebra of the gauge group7 . This gives a total of (21 + 15 + 96 = 132) scalar ﬁelds. It turns out that these scalars can be represented by a 28 × 28 matrix valued ﬁeld M satisfying8 M LM T = L, where L = I6 MT = M , −I16 · (4.16) (4.15)

I6

In denotes an n × n identity matrix. We shall choose a convention in which M = I√ corresponds to a compactiﬁcation on (S 1 )6 with each S 1 having 28 radius α = 1 measured in the string metric, and without any background gauge or antisymmetric tensor ﬁelds. We can get another scalar ﬁeld a by dualizing the gauge invariant ﬁeld strength H of the antisymmetrix tensor ﬁeld through the relation: √ H µνρ = −( −g)−1 e2Φ µνρσ ∂σ a , (4.17) where Φ denotes the four dimensional dilaton and gµν denotes the (3 + 1) dimensional canonical metric deﬁned in equations (4.1), (4.2) respectively. It is convenient to combine the dilaton Φ and the axion ﬁeld a into a single complex scalar λ: λ = a + ie−Φ ≡ λ1 + iλ2 . (4.18)

7 Only the sixteen gauge ﬁelds in the Cartan subalgebra of the gauge group can develop vacuum expectation value since such vacuum expectation values do not generate any ﬁeld strength, and hence do not generate energy density. 8 For a review of this construction, see [1].

A. Sen: Duality Symmetries

265

At a generic point in the moduli space, where the scalars M take arbitrary vacuum expectation values, the non-abelian gauge symmetry of the ten dimensional theory is broken to its abelian subgroup U (1)16 . Besides these sixteen U (1) gauge ﬁelds we get twelve other U (1) gauge ﬁelds from components Gmµ , Bmµ (4 ≤ m ≤ 9, 0 ≤ µ ≤ 3) of the metric and the anti-symmetric tensor ﬁeld respectively. Let us denote these 28 U (1) gauge ﬁelds (after suitable normalization) by Aa (1 ≤ a ≤ 28). In terms of these µ ﬁelds, the low energy eﬀective action of the theory is given by [1,30–32,34]9 , S = ¯ √ ∂µ λ∂ν λ 1 µν 1 + g Tr(∂µ M L∂ν M L) d4 x −g R − g µν 2π 2(λ2 )2 8 1 1 a b a b − λ2 g µµ g νν Fµν (LM L)ab Fµ ν + λ1 g µρ g νσ Fµν Lab Fρσ , 4 4 (4.19)

a where Fµν is the ﬁeld strength associated with Aa , R is the Ricci scalar. µ and

F aµν =

1 √ ( −g)−1 2

µνρσ

a Fρσ .

(4.20)

This action is invariant under an O(6,22) transformation10 : M → ΩM ΩT , where Ω satisﬁes: ΩLΩT = L . (4.22) Aa → Ωab Ab , µ µ gµν → gµν , λ → λ, (4.21)

An O(6,22;Z) subgroup of this can be shown to be a T -duality symmetry of the full string theory [27]. This O(6,22;Z) subgroup can be described as follows. Let Λ28 denote a twenty eight dimensional lattice obtained by taking the direct sum of the twelve dimensional lattice of integers, and the sixteen dimensional root lattice of SO(32)11 . O(6,22;Z) is deﬁned to be the subset of O(6,22) transformations which leave Λ28 invariant, i.e. acting on any vector in Λ28 , produces another vector in Λ28 . It will be useful

9 The normalization of the gauge ﬁelds used here diﬀer from that in reference [1] by a factor of two. Also there we used α = 16 whereas here we are using α = 1. 10 O(p, q) denotes the group of Lorentz transformations in p space-like and q time-like dimensions. (These have nothing to do with physical space-time, which always has only one time-like direction.) O(p, q; Z) denotes a discrete subgroup of O(p, q). 11 More precisely we have to take the root lattice of Spin(32)/Z which is obtained 2 by adding to the SO(32) root lattice the weight vectors of the spinor representations of SO(32) with a deﬁnite chirality.

266

Unity from Duality: Gravity, Gauge Theory and Strings

for our future reference to undertstand why only an O(6,22;Z) subgroup of the full O(6,22) group is a symmetry of the full string theory. Since O(6,22;Z) is a T -duality symmetry, this question can be answered within the context of perturbative string theory. The point is that although at a generic point in the moduli space the massless string states do not carry any charge, there are massive charged states in the spectrum of full string theory. Since there are 28 charges associated with the 28 U (1) gauge ﬁelds, a state can be characterized by a 28 dimensional charge vector. With appropriate normalization, this charge vector can be shown to lie in the lattice Λ28 , i.e. the charge vector of any state in the spectrum can be shown to be an element of the lattice Λ28 . Since the O(6,22) transformation acts linearly on the U (1) gauge ﬁelds, it also acts linearly on the charge vectors. As a result only those O(6,22) elements can be genuine symmetries of string theory which preserve the lattice Λ28 . Any other O(6,22) element, acting on a physical state in the spectrum, will take it to a state with charge vector outside the lattice Λ28 . Since such a state does not exist in the spectrum, such an O(6,22) transformation cannot be a symmetry of the full string theory. In order to see a speciﬁc example of a T -duality transformation, let us consider heterotic string theory compactiﬁed on (S 1 )6 with one of the circles having radius R measured in the string metric, and the rest having unit radius. Let us also assume that there is no background gauge or antisymmetric tensor ﬁelds. Using the convention of reference [1] one can show that for this background −2 R I5 . (4.23) M (H) = R2 I5 I16 Consider now the O(6,22;Z) transformation with the matrix: 0 1 I5 . Ω= 1 0 I21 Using equation (4.21) we see that this transforms M (H) to 2 R I5 (H) −2 . R = M I5 I16

(4.24)

(4.25)

A. Sen: Duality Symmetries

267

Thus the net eﬀect of this transformation is R → R−1 . It says that the heterotic string theory compactiﬁed on a circle of radius R is equivalent to the same theory compactiﬁed on a circle of radius R−1 . For this reason R = 1 √ (i.e. R = α ) is known as the self-dual radius. Other O(6,22;Z) transformations acting on (4.23) will give rise to more complicated M (H) corresponding to a conﬁguration with background gauge and/or anti-symmetric tensor ﬁelds. Besides this symmetry, the equations of motion derived from this action can be shown to be invariant under an SL(2, R) transformation of the form [30, 35, 36]

a a b Fµν → (rλ1 + s)Fµν + rλ2 (M L)ab Fµν ,

λ→

pλ + q , rλ + s (4.26)

gµν → gµν ,

M →M,

where p, q, r, s are real numbers satisfying ps − qr = 1. The existence of such symmetries (known as hidden non-compact symmetries) in this and in other supergravity theories were discovered in early days of supergravity theories and in fact played a crucial role in the construction of these theories in the ﬁrst place [30, 113]. Since this SL(2,R) transformation mixes the gauge ﬁeld strength with its Poincare dual, it is an electric-magnetic duality transformation. This leads to the conjecture that a subgroup of this continuous symmetry group is an exact symmetry of string theory [1,36–42]. One might wonder why the conjecture refers to only a discrete subgroup of SL(2, R) instead of the full SL(2, R) group as the genuine symmetry group. This follows from the same logic that was responsible for breaking O(6,22) to O(6,22;Z); however since the SL(2, R) transformation mixes electric ﬁeld with magnetic ﬁeld, we now need to take into account the quantization of magnetic charges. We have already described the quantization condition on the electric charges. Using the usual Dirac-Schwinger-Zwanziger rules one can show that in appropriate normalization, the 28 dimensional magnetic charge vectors also lie in the same lattice Λ28 . Also with this normalization convention the electric and magnetic charge vectors transform as doublet under the SL(2, R) transformation; thus it is clear that the subgroup of SL(2, R) that respects the charge quantization condition is SL(2, Z). An arbitrary SL(2, R) transformation acting on the quantized electric and magnetic charges will not give rise to electric and magnetic charges consistent with the quantization law. This is the reason behind the conjectured SL(2, Z) symmetry of heterotic string theory on T 6 . Note that since this duality acts non-trivially on the dilaton and hence the string coupling, this is a non-perturbative symmetry, and cannot be veriﬁed order by order in perturbation theory. Historically, this is the ﬁrst example of a concrete

268

Unity from Duality: Gravity, Gauge Theory and Strings

duality conjecture in string theory. Later we shall review other tests of this duality conjecture. 4.3 Duality between heterotic on T 4 and type IIA on K3 The massless bosonic ﬁeld content of heterotic string theory compactiﬁed on T 4 can be found in a manner identical to that in heterotic string theory on T 6 . Besides the dilaton Φ(H) , we get many other massless scalar ﬁelds from the internal components of the metric, the anti-symmetric tensor ﬁeld and the gauge ﬁelds. In this case these scalars can be represented by a 24 × 24 matrix valued ﬁeld M (H) satisfying M (H) LM (H)T = L, where L = I4 I4 −I16 . (4.28) M (H)T = M (H) , (4.27)

We again use the convention that M (H) = I24 corresponds to compactiﬁca√ tion on (S 1 )4 with each S 1 having self-dual radius ( α = 1), without any background gauge ﬁeld or anti-symmetric tensor ﬁeld. At a generic point in the moduli space, where the scalars M (H) take arbitrary vacuum expectation values, we get a U (1)24 gauge group, with 16 gauge ﬁelds coming from the Cartan subalgebra of the original gauge group in ten dimensions, and eight other gauge ﬁelds from components Gmµ , Bmµ (6 ≤ m ≤ 9, 0 ≤ µ ≤ 5) of the metric and the anti-symmetric tensor ﬁeld respectively. Here xm denote the compact directions, and xµ denote the non-compact directions. Let us (H)a (H) (1 ≤ a ≤ 24). Finally, let gµν denote these 24 U (1) gauge ﬁelds by Aµ (H) and Bµν denote the canonical metric and the anti-symmetric tensor ﬁeld respectively. In terms of these ﬁelds, the low energy eﬀective action of the theory is given by, SH = 1 1 d6 x −g (H) R(H) − g (H)µν ∂µ Φ(H) ∂ν Φ(H) (2π)3 2 1 µν + g Tr(∂µ M (H) L∂ν M (H) L) 8 (H) 1 (H)b (H)a − e−Φ /2 g (H)µµ g (H)νν Fµν (LM (H) L)ab Fµ ν 4 (H) 1 (H) (H) (4.29) − e−Φ g (H)µµ g (H)νν g (H)ρρ Hµνρ Hµ ν ρ , 12

**A. Sen: Duality Symmetries
**

(H)a (H)a

269

where Fµν is the ﬁeld strength associated with Aµ , R(H) is the Ricci (H) (H) scalar, and Hµνρ is the ﬁeld strength associated with Bµν :

(H) Hµνρ =

1 (H) (H)b ∂µ Bνρ + A(H)a Lab Fνρ 2 µ

+ (cyclic permutations of µ, ν, ρ). (4.30)

**This action is invariant under an O(4,20) transformation: M (H) → ΩM (H) ΩT ,
**

(H) (H) Bµν → Bµν ,

A(H)a → Ωab A(H)b , µ µ

(H) (H) gµν → gµν ,

Φ(H) → Φ(H) ,

(4.31)

where Ω satisﬁes: ΩLΩT = L . (4.32)

Again as in the case of T 6 compactiﬁcation, only an O(4,20;Z) subgroup of this which preserves the charge lattice Λ24 is an exact T -duality symmetry of this theory. The lattice Λ24 is obtained by taking the direct sum of the 8 dimensional lattice of integers and the root lattice of Spin(32)/Z2. Let us now turn to the spectrum of massless bosonic ﬁelds in type IIA string theory on K3. In ten dimensions the massless bosonic ﬁelds in type IIA string theory are the metric gMN , the rank two anti-symmetric tensor BMN and the scalar dilation Φ coming from the NS sector, and a gauge ﬁeld AM and a rank three antisymmetric tensor ﬁeld CMNP coming from the RR sector. The low energy eﬀective action of this theory involving the massless bosonic ﬁelds is given by [90] SIIA = √ 1 1 d10 x −g R − g µν ∂µ Φ∂ν Φ 7 (2π) 8 1 −Φ/2 µµ νν ρρ 1 − e g g g Hµνρ Hµ ν ρ − e3Φ/4 g µµ g νν Fµν Fµ ν 12 4 1 Φ/4 µµ νν ρρ σσ − e g g g g Gµνρσ Gµ ν ρ σ 48 1 √ (4.33) − ( −g)−1 εµ0 ···µ9 Bµ0 µ1 Gµ2 ···µ5 Gµ6 ···µ9 , (48)2

where R is the Ricci scalar, and Fµν Hµνρ Gµνρ = = = ∂µ Aν − ∂ν Aµ , ∂µ Bνρ + cyclic permutations of µ, ν, ρ , ∂µ Cνρσ + Aµ Hνρσ + (−1)P · cyclic permutations , (4.34)

are the ﬁeld strengths associated with Aµ , Bµν and Cµνρ respectively. Upon compactiﬁcation on K3 we get a new set of scalar ﬁelds from the Kahler and

270

Unity from Duality: Gravity, Gauge Theory and Strings

complex structure moduli of K3. These can be regarded as deformations of the metric and give a total of 58 real scalar ﬁelds. We get 22 more scalar ﬁelds φ(p) by decomposing the antisymmetric tensor ﬁeld BMN along the (p) twenty two harmonic two forms ωmn in K3:

22

Bmn (x, y) ∼

p=1

(p) φp (x)ωmn (y) + · · · .

(4.35)

Here {xµ } and {y m } denote coordinates along the non-compact and K3 directions respectively. These eighty scalar ﬁelds together parametrize a coset O(4, 20)/O(4) × O(20) and can be described by a matrix M (A) satisfying properties identical to those of M (H) described in (4.27). This theory also has twenty four U (1) gauge ﬁelds. 22 of the gauge ﬁelds arise from the components of the three form ﬁeld CMNP :

22

Cmnµ (x, y) =

p=1 (p)

(p) ωmn (y)A(p) (x) + . . . . µ

(4.36)

Aµ deﬁned in (4.36) behaves as gauge ﬁelds in six dimensions. One more gauge ﬁeld comes from the original RR gauge ﬁeld Aµ . The last one Aµ comes from dualizing Cµνρ : G∼

∗

(dA) ,

(4.37)

where ∗ denotes Poincare dual in six dimensions. Together we shall denote (A)a for 1 ≤ a ≤ 24. Besides these ﬁelds, the theory these gauge ﬁelds by Aµ contains the canonical metric and the anti-symmetric tensor ﬁeld which we (A) (A) shall denote by gµν and Bµν respectively. The action involving these ﬁelds is given by, SA = 1 1 d6 x −g (A) R(A) − g (A)µν ∂µ Φ(A) ∂ν Φ(A) 3 (2π) 2 1 µν + g Tr(∂µ M (A) L∂ν M (A) L) 8 1 (A) (A)b (A)a − eΦ /2 g (A)µµ g (A)νν Fµν (LM (A) L)ab Fµ ν 4 (A) 1 (A) (A) − e−Φ g (A)µµ g (A)νν g (A)ρρ Hµνρ Hµ ν ρ 12 1 (A) (A)a (4.38) − εµνρδ η ( −g (A) )−1 Bµν Fρδ Lab F (A)b , η 16

(A)a

where Fµν is the ﬁeld strength associated with Aµ , R(A) is the Ricci (A) (A) scalar, and Hµνρ is the ﬁeld strength associated with Bµν :

(A) (A) Hµνρ = ∂µ Bνρ + (cyclic permutations of µ, ν, ρ) .

(A)a

(4.39)

A. Sen: Duality Symmetries

271

In writing down the above action we have used the convention that M (A) = I24 corresponds to compactiﬁcation on a speciﬁc reference K3, possibly with speciﬁc background Bmn ﬁelds. This action has an O(4,20) symmetry of the form: M (A) → ΩM (A) ΩT ,

(A) (A) Bµν → Bµν ,

A(A)a → Ωab A(A)b , µ µ

(A) (A) gµν → gµν ,

Φ(A) → Φ(A) , ΩLΩT = L .

(4.40)

where Ω satisﬁes: (4.41)

An O(4,20;Z) subgroup of this can be shown to be an exact T -duality symmetry of string theory [126]. The lattice Λ24 which is preserved by this O(4,20;Z) subgroup of O(4,20) is not the lattice Λ24 deﬁned earlier, but is in general an O(4,20) rotation of that lattice: Λ24 = Ω0 Λ24 . (4.42)

Ω0 depends on the choice of the special reference K3 mentioned earlier. It is now a straightforward exercise to show that the equations of motion and the Bianchi identities derived from (4.29) and (4.38) are identical if we use the following map between the heterotic and the type II variables [23,43]:

(H) (A) gµν = gµν ,

M (H) = ΩM (A) ΩT ,

Φ(H) = −Φ(A) ,

A(H)a = Ωab A(A)a , µ µ 1 µνρδ η (A) (H) (H)µνρ −g (H) exp(−Φ )H = ε Hδ η . 6

(4.43)

where Ω is an arbitrary O(4,20) matrix. This leads to the conjectured equivalence between heterotic string theory compactiﬁed on T 4 and type IIA string theory compactiﬁed on K3 [23]. But clearly the two theories cannot be equivalent for all Ω since in the individual theories the O(4,20) symmetry is broken down to O(4,20;Z). Ω can be found (up to an O(4,20;Z) transformation) by comparing the T -duality symmetry transformations in the two theories. To do this let us note that according to equation (4.43) a transformation M (H) → ΩM (H) ΩT will induce a transformation M (A) → (Ω−1 ΩΩ)M (A) (Ω−1 ΩΩ)T . (4.44) Thus if Ω preserves the lattice Λ24 , Ω−1 ΩΩ should preserve the lattice Λ24 = Ω0 Λ24 . This happens if we choose: Ω = Ω−1 . 0 (4.45) Note again that there is a relative minus sign that relates Φ(H) and Φ(A) , showing that the strong coupling limit of one theory corresponds to the weak coupling limit of the other theory.

s to be integers instead of arbitrary real numbers.R) transformation [44]: λ→ gµν pλ + q Bµν p . and a rank four anti-symmetric tensor ﬁeld Dµνρσ whose ﬁeld strength is self-dual. or by explicit computation in string theory. The RR sector contributes a scalar ﬁeld a sometimes called the axion.R). ps − qr = 1 .4 SL(2 . non-renormalization theorems tell us that this is exact to this order in the space-time derivatives. r. and a scalar ﬁeld Φ known as the dilaton. which are in fact just the equations of motion of type IIB supergravity. The NS sector gives the graviton described by the metric gµν .1.Z) subgroup of this SL(2. q s Bµν . s are real numbers satisfying. (4. .2 for heterotic string theory on T 6 .Z) duality transformations in the two theories. Actually it turns out that there is no simple covariant action for this low energy theory.46) The low energy eﬀective action in this theory can be determined either from the requirement of supersymmetry. although both transform as modulus under the respective SL(2. Gauge Theory and Strings 4. and so there is no scope for the quantum corrections to change the form of the action.48) The existence of this SL(2. but there are covariant ﬁeld equations [44]. obtained by restricting p. r. an anti-symmetric tensor ﬁeld Bµν . It is often convenient to combine the axion and the dilaton into a complex scalar ﬁeld λ as follows12 : λ = a + ie−Φ/2 ≡ λ1 + iλ2 . Basically supersymmetry determines the form of the equations of motion to this order in the derivatives completely. (4. another rank two anti-symmetric tensor ﬁeld Bµν . The main point is that these equations of motion are covariant (in the sense that they transform into each other) under an SL(2. Dµνρσ → Dµνρσ . Z ) self-duality of type IIB in D = 10 As described in Section 2.272 Unity from Duality: Gravity. q. the massless bosonic ﬁelds in type IIB string theory come from two sectors. is 12 Note that this ﬁeld λ has no relation to the ﬁeld λ deﬁned in Section 4. q. Bµν (4. Although in string theory this low energy theory is derived from the tree level analysis.47) where p. we shall not explicitly write down the equations of motion.R) symmetry in the type IIB supergravity theory led to the conjecture that an SL(2. − Neveu-Schwarz−Neveu-Schwarz (NS) and Ramond-Ramond (RR). → Bµν r rλ + s → gµν . For the sake of brevity.

Bµν and Φ invariant. but leaves gµν . Thus p must be an integer.Z) is a symmetry of string theory. the SL(2. Z) matrices r s with p. R) map q trices consistent with charge quantization are the SL(2. but does not prove that SL(2. In particular. q. Note that this argument only shows that SL(2. it cannot be veriﬁed order by order in string perturbation theory. • Ω: this is the world-sheet parity transformation mentioned in Section 2.Z) transformation laws of the .Z) is the maximal possible subgroup of SL(2. and α a ﬁxed constant.R) to SL(2. whose vacuum expectation value represents the string coupling constant. s integers satisfying ps − qr = 1. r. s integers satisfying (ps − qr) = 1. it carries exactly one unit of Bµν charge. it changes the sign of Bµν .R) that can be a symmetry of the full string theory. as the spectrum of string theory does not contain fractional strings carrying a fraction of the charge carried by the elementary string.R) for which p is always an integer consists of matrices of the form p α−1 r αq s . it changes the sign of a. type IIB theory has two perturbatively veriﬁable discrete Z2 symmetries.R) transformation gives a state with p units of Bµν charge and r units of Bµν charge.Z) transformation. This means that the Bµν charge must be quantized in integer units. acting on the massless bosonic sector ﬁelds. The breaking of SL(2. (4.47) we see that acting on an elementary string state carrying one unit of Bµν charge. r. we see that the eﬀect of (−1)FL · Ω is to change of sign of Bµν and Bµν . Acting on the massless bosonic sector ﬁelds.and the right-moving sectors of the world-sheet. leaving the other ﬁelds invariant. Bµν and Dµνρσ . An elementary string is known to carry Bµν charge. It is easy to see that the maximal subgroup of SL(2. From (4. They are as follows: • (−1)FL : it changes the sign of all the Ramond sector states on the left moving sector of the world-sheet. Absorbing α into a redeﬁnition of Bµν we see that the subgroup of SL(2. since SL(2. Besides this non-perturbative SL(2.Z) can be seen as follows. q. leaving the other massless bosonic ﬁelds invariant. In suitable normalization convention.49) with p.Z) acts non-trivially on the dilaton. From this description.A. We shall see later how one can ﬁnd non-trivial evidence for this symmetry. Comparing this with the action of the SL(2. a and Dµνρσ . Sen: Duality Symmetries 273 a symmetry of the full string theory [23].1 that exchanges the left. In particular.

The open string states of type I string theory are the analogs of twisted sector states in an orbifold. 4. 5. Z) × SL(3.274 Unity from Duality: Gravity.50) This information will be useful to us later. Z) SO(7. (4. Theories obtained by modding out (compactiﬁed) type IIB string theory by a discrete symmetry group. Z) Note that besides the full duality group. we have also displayed the T duality group of each theory which can be veriﬁed order by order in string . Gauge Theory and Strings massless bosonic sector ﬁelds. As mentioned earlier. Z) SO(6. 3. We now list the conjectured self-duality groups of type IIA/IIB string theory compactiﬁed on T n for diﬀerent values of n [23]: D = (10 − n) 9 8 7 6 5 4 3 2 Full Duality Group SL(2. We begin with the self duality groups of type II string theories compactiﬁed on tori of diﬀerent dimensions. This corresponds to type I string theory. 6. Z) SO(3. Thus for n ≥ 1. Here we shall list the main series of such duality conjectures. Z) SL(2. 8. Z) SO(4. Z) SO(8. which must be added to the theory in order to maintain ﬁniteness. Z) × SL(2.5 Other examples Following the same procedure. the self-duality groups of type IIA and type IIB theories compactiﬁed on an n-dimensional torus T n will be identical. namely. The simplest example of an orientifold is type IIB string theory modded out by Ω. Z) SO(5. where some of the elements of the group involve Ω. 5. are known as orientifolds [100. 7. we are led to many other duality conjectures in theories with 16 or more supersymmetry generators. Z) E6(6) (Z) E7(7) (Z) E8(8) (Z) E8(8) (Z) T -duality Group − SL(2. 4. we see that (−1)FL · Ω can be identiﬁed with the SL(2.Z) transformation: −1 −1 . Z) SL(5. 101]. Z) SO(5. there is a T -duality that relates type IIA on a circle to type IIB on a circle of inverse radius. studying symmetries of the eﬀective action together with charge quantization rules. The closed string sector of type I theory consists of the Ω invariant states of type IIB string theory.

21. 21. 20. Z) Since type I and SO(32) heterotic string theories are conjectured to be dual to each other in ten dimensions.Z) subgroup of this is an exact self-duality symmetry of the type IIB string . in the case of type I string theory. and were known as hidden non-compact symmetries. Z) × SL(2. Z) O(8. Z) T -duality Group O(1. Z) O(4. 23. Z) O(3. Z) O(3.21) symmetry [43]. 19. Z) O(8. 7. and En(n) (Z) denotes a discrete subgroup of En(n) . the two heterotic string theories can be shown to be related by a T -duality transformation. but charge quantization breaks this symmetry to its discrete subgroup. 18. 19. upon compactiﬁcation on T n . both of them will have the same self-duality group. As a result. 22. 18. Z) O(6. Z) O(7. there is no perturbatively realised self-duality group (except trivial transformations which are part of the SO(32) gauge group and the group of global diﬀeomorphisms of T n ). Z) O(6. 24. Although there are two distinct heterotic string theories in ten dimensions.A. the low energy eﬀective ﬁeld theory is invariant under the full continuous group [45]. En(n) denotes a non-compact version of the exceptional group En for n = 6. Z) O(8. Z) O(2. 24. Sen: Duality Symmetries 275 perturbation theory. We now display this self-duality group in various dimensions: D = (10 − n) 9 8 7 6 5 4 3 2 Full Duality Group O(1. 22. In this case one expects a very large duality symmetry group based on hyperbolic Lie algebra [103]. upon compactiﬁcation on a circle. which is not well understood to this date. Z) O(2. Z) O(5. 17. Note that we have stopped at D = 2. 8. these symmetries were discovered in the early days of supergravity theories. However. 17. As noted before. 20. Next we turn to the self-duality conjectures involving compactiﬁed heterotic string theories. Z) O(4. the second column of the above table also represents the duality symmetry group of type I string theory on T n . G for any group G denotes the loop group of G based on the corresponding aﬃne algebra and G(Z) denotes a discrete subgroup of this loop group. We could in principle continue this all the way to D = 1 where all space-like directions are compactiﬁed. In each of the cases mentioned. 24. which leads to the conjecture that an SO(5.21. The eﬀective action of type IIB string theory compactiﬁed on K3 has an SO(5. Z) O(5.

Thus for example . But once we have arrived at a duality conjecture based on the analysis of the low energy eﬀective action. provides us with only a crude test of duality. 5 Precision test of duality: Spectrum of BPS states Analysis of the low energy eﬀective action. type IIA/IIB on K3 × T n are dual to heterotic string theory on T n+4 for n ≥ 1. Thus the self-duality symmetry groups in these theories can be read out from the second column of the previous table displaying the self-duality groups of heterotic string theory on T n . we can perform a much more precise test by analysing the spectrum of BPS states in the theories. where massless states form a shorter representation of the algebra than massive states. Quite often the eﬀective action contains another set of terms satisfying non-renormalization theorems. and are known as Green-Schwarz terms. Later we shall also describe a more systematic way of “deriving” various duality conjectures from some basic set of dualities. but I shall not discuss it in this article. due to the conjectured duality between type IIA on K3 and heterotic on T 4 . BPS states are states which are invariant under part of the supersymmetry transformation. The duality symmetry groups of these theories can again be guessed from an analysis of the low energy eﬀective ﬁeld theory and the charge quantization conditions. This has an analog in the theory of representations of the Lorentz group. Indeed. Finally. The conjectured duality between type IIA string theory compactiﬁed on K3 and heterotic string theory compactiﬁed on T 4 has already been discussed before. as discussed in the last section. Adler-Bardeen theorem guarantees that they are not renormalized beyond one loop. and are characterized by two important properties: • They belong to a supermultiplet which has typically less dimension than a non-BPS state. Besides the theories discussed here. most of the duality conjectures in string theory were arrived at by analysing the symmetries of the low energy eﬀective action. type IIA on K3 × T n is equivalent to type IIB on K3 × T n . there are other theories with 16 or more supercharges obtained from non-geometric compactiﬁcation of heterotic/type II string theories [46–48].276 Unity from Duality: Gravity. Gauge Theory and Strings theory on K3. They are required for anomaly cancellation. Although in this section I have focussed on duality symmetries of the low energy eﬀective action which satisfy a non-renormalization theorem as a consequence of space-time supersymmetry. this is not the only part of the full eﬀective action which satisfy such a non-renormalization theorem. Its value lies in its simplicity. These terms have also been used eﬀectively for testing various duality conjectures [127]. Due to the equivalence of type IIB on S 1 and type IIA on S 1 .

and {y} denotes the coordinates labelling the moduli space of the theory13 . (5. Thus in this basis the supersymmetry algebra has the form: {Qα . for 1 ≤ α. Q. In this case. appear in the algebra. fαβ has no zero eigenvalue. e. (5. which is 2N/2 dimensional. Here m denotes the rest mass of the particle.g. In this case by taking appropriate linear combinations of Qα we can diagonalize f . (We are considering the case where N is even.A. Qβ } = δαβ .3) We can form an irreducible representation of this algebra by taking all states to be annihilated by Qα for α > M . by taking linear combinations of the Qα we can bring the algebra into the form: {Qα . known as central charges. • The mass of a BPS state is completely determined by its charge as a consequence of the supersymmetry algebra. f has (N − M ) zero eigenvalues for some M < N .2) This is the N dimensional Cliﬀord algebra. Q denotes various gauge charges carried by the particle. We shall now consider the following distinct cases: 1. This statement also has an analog in the theory of representations of the Lorentz algebra. the supersymmetry algebra takes the form: {Qα . Sen: Duality Symmetries 277 a photon has only two polarizations but a massive vector particle has three polarizations. This relation between the mass and the charge is known as the BPS mass formula. a spin 1 representation of the Lorentz algebra containing only two states must be necessarily massless. {y}) . Qβ } = fαβ (m. we can bring f into the identity matrix. By a further appropriate rescaling of Qα . . 2. β ≤ M . 0 for α or β > M . Suppose the theory has N real supersymmetry generators Qα (1 ≤ α ≤ N ). Acting on a single particle state at rest.1) where fαβ is a real symmetric matrix which is a function of its arguments m. Q and {y}. In that case the states will 13 Only speciﬁc combinations of Q and {y}. Qβ } = = δαβ .) Such states would correspond to non-BPS states. Thus the single particle states under consideration form a representation of this Cliﬀord algebra. We shall now explain the origin of these two properties [29]. (5.

i. Since M < N . This theory also has BPS states breaking half the space-time supersymmetry. these states are invariant under part of the supersymmetry algebra generated by Qα for α > M . The total number of supersymmetry generators in this theory is 32. As another example. We can get diﬀerent kinds of BPS states depending on the value of M . Gauge Theory and Strings form a representation of an M dimensional Cliﬀord algebra generated by Qα for 1 ≤ α ≤ M . and are known as short multiplets. and is the origin of the BPS formula relating the mass and the charge of the particle.e. From this discussion it is clear that in order to get a BPS state. We can also have BPS states breaking 3/4 of the space-time supersymmetry (M = 24). We shall discuss this relation as well as the origin of these BPS states in more detail later. . These are known as long multiplets. These are known as BPS states. we see that these are lower dimensional representations compared to that of a generic non-BPS state. the matrix f must have some zero eigenvalues. We can also have states that break 3/4 of the supersymmetries14 . Consider type IIB string theory compactiﬁed on a circle S 1 . This constraint arises due to the fact that the unbroken supersymmetry generators must form a representation of the little group SO(9 − n) of a massive particle in (10 − n) dimensional space-time. depending on the number of supersymmetry generators that leave the state invariant. This in turn. These states are known as ultra-short multiplets. In each case there is a speciﬁc relation between the mass and the various charges carried by the state. These will form a 212 = 256 × 16 dimensional representation. This representation is 2M/2 dimensional for M even. then they belong to a 24 = 16 dimensional representation of the supersymmetry algebra. Furthermore. BPS states are further characterized by the property that the degeneracy of BPS states with a given set of charge quantum numbers is independent of the value of the moduli ﬁelds {y}. charges Q and the moduli {y}.278 Unity from Duality: Gravity. For these states M = 16 and hence we have 28 = 256 dimensional representation of the supersymmetry algebra. consider heterotic string theory compactiﬁed on an n-dimensional torus T n . But if we consider states that are invariant under half of the supercharges. These belong to a 64 dimensional representation of the supersymmetry algebra known as intermediate states. gives a constraint involving mass m. Thus a generic non-BPS state will belong to a 28 = 256 dimensional representation of the supersymmetry algebra. This is known as the short representation of this superalgebra. Since string coupling is also one of 14 It turns out that these states can exist only for n ≥ 5. let us illustrate the preceeding discussion in the context of a string theory. Thus a generic non-BPS supermultiplet is 216 = (256)2 dimensional. Before we proceed. The original theory has 16 supercharges.

Sen: Duality Symmetries 279 the moduli of the theory. so let us review the argument leading to this property [29]. but with quantum numbers that are not present in the spectrum of elementary string states. We shall now illustrate this procedure with the help of speciﬁc examples. 65]. Identify BPS states in the spectrum of elementary string states. 2. but it can be applied to any other theory. . we require that the mass of the BPS state should stay away from the continuum. we can now adapt the following strategy to carry out tests of various duality conjectures using the spectrum of BPS states in the theory: 1. then this is clearly not possible since other multiplets have diﬀerent number of states. Thus these states must arise as solitons/composite states. this implies that the degeneracy at any value of the string coupling is the same as that at weak coupling. The spectrum of these BPS states can be trusted at all values of the coupling even though it is calculated at weak coupling. Given this result. This typically takes a BPS state in the spectrum of elementary string states to another BPS state. as long as it stays ultra-short. We shall discuss this in the context of the speciﬁc example of type IIB string theory compactiﬁed on S 1 . an ultra-short multiplet stays ultra-short as we move in the moduli space [88]. This is the key property of the BPS states that makes them so useful in testing duality. Note that for this argument to be strictly valid. since otherwise the counting of states is not a well deﬁned procedure. Try to explicitly verify the existence of these solitonic states with degeneracy as predicted by duality. We shall mainly follow [51. Furthermore.A. Suppose the theory has an ultra-short multiplet at some point in the moduli space. Now let us change the moduli. This requires that the mass of a BPS state should be strictly less than the total mass of any set of two or more particles carrying the same total charge as the BPS state. Make a conjectured duality transformation. This will provide a non-trivial test of the corresponding duality conjecture. 3. 62. Thus as long as the spectrum varies smoothly with the moduli (which we shall assume). its mass is determined by the BPS formula. Thus we see that the degeneracy of ultra-short multiplets cannot change as we change the moduli of the theory. A similar argument can be given for other multiplets as well. The question that we shall be asking is: can the ultrashort multiplet become a long (or any other) multiplet as we change the moduli? If we assume that the total number of states does not change discontinuously.

As discussed in Section 4. Z) duality of the heterotic string theory will require the N = 4 supersymmetric SU (2) gauge theory to have this SL(2. scalar self-interaction etc. This can be done for example by taking a particular S 1 in T 6 to be orthogional to all other circles. taking the components of the gauge ﬁelds along this S 1 to be zero. together with a set of decoupled N = 4 supersymmetric U (1) gauge theories and N = 4 supergravity.1 SL(2 . The form of the Lagrangian is ﬁxed completely by the requirement of N = 4 supersymmetry up to two independent parameters − the coupling constant g that determines the strength of all interactions (gauge. we get unbroken E8 × E8 or SO(32) gauge symmetry. Z) symmetry of heterotic string theory. 50]. Thus by testing the duality invariance of the spectrum of this N = 4 supersymmetric SU (2) gauge theory we can test the conjectured SL(2. Z ) S-duality in heterotic on T 6 and multi-monopole moduli spaces As discussed in Section 4. The conjectured SL(2. The N = 4 supersymmetric SU (2) gauge theory has a vector. at a generic point in the moduli space the unbroken gauge group is U (1)28 . In that case the eﬀective ﬁeld theory at energies much below the string scale will be described by an N = 4 supersymmetric SU (2) gauge theory.Z) duality symmetry.280 Unity from Duality: Gravity. we can analyse the spectrum near some particular point in the moduli space. Yukawa.18) through the relation: λ = 4π θ +i 2 · 2π g (5. and the vacuum angle θ that multiplies the topological term Tr(F F ) involving the gauge ﬁeld. . g and θ are related to the vacuum expectation value of the ﬁeld λ deﬁned in (4.4) 15 Independently of string theory. Gauge Theory and Strings 5. But there are special points in this moduli space where we get enhanced non-abelian gauge group [107].2.2. Since the BPS spectrum does not change as we change the moduli. Z) symmetry15 . if we set the internal components of the original ten dimensional gauge ﬁelds to zero.). Let us consider a special point in the moduli space where an SU (2) gauge symmetry is restored. Thus for example. heterotic string theory compactiﬁed on T 6 is conjectured to have an SL(2. and taking the radius of this S 1 to be the self-dual radius. six massless scalars and four massless Majorana fermions in the adjoint representation of SU (2) [50]. With the choice of suitable normalization convention. In this subsection we shall see how one can test this conjecture by examining the spectrum of BPS states. the existence of a strong-weak coupling duality in this theory was conjectured earlier [49.

When SU (2) is broken to U (1) by the vacuum expectation value of φm . the electric charge quantum number ne and the magnetic charge quantum number nm .6) does not break supersymm metry. (5. In this notation the elementary W + nm 1 boson corresponds to a state. The parameters {am } correspond to the vacuum expectation values of a subset of the scalar moduli ﬁelds M in the full string theory. so that gravity is still decoupled from this gauge theory. the spectrum of solitons in this theory is characterized by two quantum numbers. we can easily work out its action ne on the charge quantum numbers [1]. (5. We ne shall denote such a state by . From the relation ps − qr = 1 satisﬁed 0 r .7) for appropriate choice of sign convention for ne and nm . m (5. Thus acting on an 1 p state it produces a state. these states become massless and form part of the SU (2) gauge multiplet.6) Vacuum expectation values of φα of the form (5. We shall work in a region in the moduli scale of breaking of SU (2) is small space where am = 0 for some m. These break half of the 16 space-time supersymmetry generators and hence form a 28/2 = 16 dimensional representation of the supersymmetry algebra. These states can be found explicitly in the spectrum of elementary string states from the sector containing strings with one unit of winding and one unit of momentum along the special S 1 that is responsible for the enhanced SU (2) gauge symmetry. By studying the action of the SL(2. 1 ≤ m ≤ 6) is proportional to ( m<n α αβγ β γ 2 φm φn ) . As we approach the point in the moduli space where this special S 1 has self-dual radius. The BPS states in the spectrum of elementary particles in this theory are the heavy charged bosons W ± and their superpartners.26) on the gauge ﬁelds. Sen: Duality Symmetries 281 The potential involving the six adjoint representation scalar ﬁelds φα (1 ≤ m α ≤ 3.5) This vanishes for φα = am δα3 . normalized so that ne and nm are both integers.A. Z) 0 transformation (4. but breaks the gauge group SU (2) to U (1). but the√ compared to the string scale (|am | << ( α )−1 ) for all m). The answer is nm ne nm → p r q s ne nm .

we can ﬁnd integers q and s satisfying ps − qr = 1. we can easily see that p and r are relatively prime. for a given r. Thus the states obtained by quantizing the bosonic sector of p the theory has charge quantum numbers for all integer p. this system has 4r bosonic collective coordinates. 1 The degeneracy comes from quantizing the fermionic sector. which describe the result of applying the eight broken supersymmetry generators on the monopole solution. States with r = 1 come from one monopole solution. and the fourth one is an angular variable describing the U (1) phase of the monopole. We can divide this set into the “center of mass” coordinates containing four bosonic and eight fermionic coordinates. correspond to the spatial location and the U (1) phase of each of the r monopoles. The theory contains classical monopole solutions which break half of the supersymmetries of the original theory. Z) duality predicts that for every p and r relatively prime. The momenta conjugate to the ﬁrst three coordinates correspond to the components of the physical momentum of the particle. These solutions are non-singular everywhere. we need to quantize these collective excitations and look for supersymmetric ground states of the corresponding quantum mechanical system. exactly as predicted by SL(2. This has four bosonic collective coordinates. Z) [50]. Furthermore for every p and r relatively prime. There are eight fermionic zero modes. The fourth coordinate is periodically identiﬁed and hence its conjugate momentum is quantized in integer units. Thus the ground state has 24 = 16-fold degeneracy. The total number of fermionic collective coordinates can be computed from an index theorem and is equal to 8r [106]. Thus SL(2. In order to study the spectrum of BPS solitons. and in fact. three of which correspond to the physical position of the monopole in the three dimensional space. Z) matrix. Each solution also has inﬁnite number of vibrational modes with non-zero frequency. when the monopoles are far away from each other. These 4r parameters correspond to the bosonic collective excitations of this system [105]. As has already been said. This integer p corresponds to the electric charge quantum number ne .282 Unity from Duality: Gravity. Let us now turn to the analysis of states with r > 1 [51]. These can be set to zero by working in the rest frame of the monopole. but excitations of these modes are not relevant for ﬁnding supersymmetric ground states. These form an eight dimensional Cliﬀord algebra. the theory must contain a unique short multiplet with p charge quantum numbers [51]. there is a 4r parameter non-singular solution with r units of total magnetic charge [52. and the “relative coordinates” containing 4(r − 1) bosonic and 8(r − 1) . r We can now directly examine the solitonic sector of the theory to check this prediction. which. Gauge Theory and Strings by an SL(2. 104].

55]. But it turns out that the eﬀect of the 8(r − 1) fermionic degrees of freedom in the quantum mechanical system makes the wave-function a diﬀerential form of arbitrary rank on Mr [54. and Mr by the relative coordinates. we need to turn to the quantum mechanics of the relative coordinates. Due to this identiﬁcation. They are listed below: • First of all. 53. 105]. The full moduli space has the structure [52]: (R3 × S 1 × Mr )/Zr . • Normally the part of the wave-function involving the relative coordinates will be a function on Mr . the ones saturating Bogomol’nyi bound correspond to harmonic diﬀerential forms on Mr . Since the part of the wave-function involving the coordinate of S 1 picks up a phase exp(2πip/r) under this Zr . known as the relative moduli space of r monopoles [52. Sen: Duality Symmetries 283 fermionic coordinates.A. It can be shown that the Hamiltonian of the . 53]. we see that the wave-function involving the relative coordinates must pick up a phase of exp(−2πip/r) under this Zr transformation. consistent with the fact that a short multiplet is 16-fold degenerate. the center of mass and the relative coordinates do not completely decouple. (5. There are several subtleties with this system. It turns out that the bosonic coordinates in the relative coordinate system describe a non-trivial 4(r − 1) dimensional manifold. p r being the momentum conjugate to the overall U (1) phase. among all the possible states. The quantization of the center of mass system gives p states carrying charge quantum numbers with 16-fold degeneracy. This shows that the degeneracy is always a multiple of 16. in order to determine the number of short multiplets for a given value of p. the total wave-function must be invariant under this Zr transformation. since the total wave-function is a product of the wave-function of the center of mass system and the relative system. At this stage p can be any integer. There is an identiﬁcation of points in the product space R3 × S 1 × Mr by a Zr transformation that acts as a shift by 2π/r on S 1 and as a diﬀeomorphism on Mr without any ﬁxed point [52. However.8) where R3 is parametrized by the center of mass location. • Finally. although they decouple locally. This can be understood as follows. The quantum mechanics of the bosonic and fermionic relative coordinates can be regarded as that of a supersymmetric particle moving in this moduli space. not necessarily prime relative to r. S 1 by the overall U (1) phase.

Also it turns out that the BPS mass formula is saturated by contribution from the center of mass coordinates. The Fock vacuum. the spectrum of elementary string states in the heterotic string theory on T 6 contains many other BPS states. Besides the BPS states discussed here. 56. For r = 2 the relevant harmonic form can be constructed explicitly [51. . In the above equation ¯ the terms in the square bracket denote the total contribution to L0 and L0 from the oscillators. Hence in order to get a BPS state.e. 2 (5. is characterized by a pair of vectors (kL . kR ) specifying the charges (momenta) associated with the six right-handed and twenty two left-handed currents on the world-sheet.57].and 16 In this and all subsequent mass formula λ should really be interpreted as the vacuum 2 expectation value of λ2 . According to the prediction of SL(2. and the vacua in the right. in turn. However general arguments showing the existence of the necessary harmonic forms has been given [58. it must be a harmonic form on Mr . therby verifying the existence of the states predicted by SL(2. a generic state is created by applying oscillators from the left. the internal momenta. and not for other values of p [51]. In the world-sheet theory. kR ) are just appropriate linear combinations of the charges carried by the state under the 28 U (1) gauge ﬁelds.and the left-moving sectors of the world-sheet. The tree level mass formula for an elementary string state in the NS sector is given by16 . Z) such a harmonic form should exist only for p and r relatively prime. Thus for every harmonic diﬀerential form we get a short multiplet. Z) duality. From the viewpoint of the spacetime theory.284 Unity from Duality: Gravity.and the right-moving sector on the Fock vacuum. the part of the wave-function involving the relative coordinates must be an eigenstate of the corresponding Hamiltonian with zero eigenvalue i. p r would require the existence of a harmonic form on Mr that picks up a phase of exp(2πip/r) under the action of Zr . 59]. Thus the existence of a short multiplet of charge quantum numbers m2 = 4 λ2 2 4 1 kR + NR − = 2 2 λ2 2 kL + NL − 1 . For r > 2 the analysis is more complicated since the metric in the multimonopole moduli space is not known. these 28 components of (kL . since the fermionic degrees of freedom associated with the center of mass coordinates supply the necessary 16-fold degeneracy. Gauge Theory and Strings relative coordinates correspond to the Laplacian on M.9) where NR and NL denote respectively the oscillator levels of the state in the right.

Thus the SL(2. (In the covariant formulation these can be traced to the contributions from the world-sheet ghost ﬁelds).11) The degeneracy d(NL ) of short multiplets for a given set of kL . this is consistent with their description as heavy gauge bosons in an N = 4 supersymmetric gauge theory. From equation (5.A. Z) self-duality symmetry of the 17 In counting degeneracy we are only counting the number of short multiplets. The next interesting class of states are the ones with NL = 1. and ignoring the trivial factor of 16 that represents the degeneracy within each short multiplet. kR is determined by the number of ways a level NL state can be created out of the Fock vacuum by the 24 left-moving bosonic oscillators (in the light-cone gauge) − 8 from the transverse bosonic coordinates of the string and 16 from the bosonization of the 32 left-moving fermions on the world-sheet − and is given by the formula: ∞ d(NL )q NL = NL =0 1 · (1 − q n )24 n=1 ∞ (5. 2 R (5. But here (and in the rest of the article) 2 we quote the ADM mass measured in the canonical metric gµν .12) The BPS states discussed earlier − the ones which can be regarded as the massive gauge bosons of a spontaneously broken non-abelian gauge theory − correspond to the NL = 0 states in this classiﬁcation. From (5. 2 (5. It turns out that of the full set of elementary string states.12) we see that they have degeneracy 2417 . Sen: Duality Symmetries 285 the left-moving sectors respectively. From equations (5. An SL(2. Z) transformation relates these states to appropriate magnetically charged states with r units of magnetic charge and p units of electric charge for p and r relatively prime.9) we see that for these states NL = 1 2 2 (k − kL ) + 1 . . Normally we do not have the factor of λ−1 in the mass formula since the formula refers to the ADM mass measured 2 in the string metric Gµν = λ−1 gµν . This is more convenient for discussing duality invariance of the spectrum.12) we see that we have only one short multiplet for states with this quantum number. The additive factor of −1/2 and −1 can be interpreted as the contributions ¯ to L0 and L0 from the vacuum. since it is gµν and not Gµν that remains invariant under a duality transformation. only those states which satisfy the constraint [64] NR = 1 .10) correspond to BPS states (short multiplets).

21. As a result. 23. both from the left. analysis based on these solutions has been of limited use. As usual we have set α = 1.and the right-moving sector of the world-sheet. (5. In this section I shall discuss the consequence of this conjectured symmetry for the spectrum of BPS states in type IIB string theory compactiﬁed on a circle S 1 . The main problem is that unlike the NL = 0 states. 26. obtained by applied appropriate oscillators. and w. Gauge Theory and Strings heterotic string theory predicts the existence of 24-fold degenerate solitonic states with these charge quantum numbers. There are inﬁnite tower of states with this quantum number. the problem has now been solved for r = 1 [70. In the world-sheet theory describing ﬁrst quantized string theory. also an integer. measured in the ten dimensional . √ 1/4 1/4 kR = (kλ2 /R + wR/λ2 )/ 2 . 72–74]. Verifying the existence of these solitonic states turns out to be quite diﬃcult [60]. k/R denotes the momentum along S 1 with k being an integer. 25. or has strong curvature at the core where the low energy approximation breaks down. to which we now turn.2 SL(2 . The spectrum of elementary string states in this theory are characterized by two charges kL and kR deﬁned as: √ 1/4 1/4 kL = (kλ2 /R − wR/λ2 )/ 2. Z) duality symmetry group. Z ) duality in type IIB on S 1 and D-branes As discussed earlier. The mass formula for any state in this tower. 61]. denotes the number of times the elementary string is wound along S 1 . Similar analysis based on soliton solutions of low energy supergravity theory has been used to test many other duality conjectures [20. Nevertheless. and one ﬁnds that these solitons have exactly the correct degeneracy 24. kL and kR denote the left and the right-moving momenta respectively.13) where R denotes the radius of S 1 measured in the ten dimensional canonical metric. the solitonic states (known as H-monopoles) which are related to the NL = 1 states by SL(2. and hence we cannot unambiguously determine the dynamics of collective coordinates of these solitons just from the low energy eﬀective ﬁeld theory. see [62. on the Fock vacuum of the world-sheet theory carrying these quantum numbers. type IIB string theory in ten dimensions has a conjectured SL(2. 5. One of the main problems with this approach has been that unlike the example discussed in this section. most of these other solutions are either singular. For details.286 Unity from Duality: Gravity. Z) duality turn out to be singular objects. The situation changed after the advent of D-branes. 63].

and the right. but here it comes due to the fact that we are using the ten dimensional canonical metric instead of the string metric to deﬁne the mass of a state.) Most of these states are not BPS states as they are not invariant under any part of the supersymmetry transformation. This requires kL = ±kR or. From equation (5. In normal convention.2). On the other hand.(right-) moving sector of the world-sheet.17) (5. 2 R (5. 18 We have stated the formula in the RR sector. a state with either NL = 0 or NR = 0 will break (3/4)th of the total number of supersymmetries in the theory. but due to space-time supersymmetry we get identical spectrum from the NS and the R sectors. It turns out that in order to be invariant under half of the space-time supersymmetry coming from the left. λ2 λ2 287 (5. is given by: 2 2 2 2 m2 = √ (kL + 2NL ) = √ (kR + 2NR ) . If. NR denote oscillator levels on the left. there will be an additional multiplicative factor of R−2/9 in the expression for m2 . then the BPS mass formula takes the form: 2k 2 m2 = √ R · λ2 NL is determined in terms of kL and kR through the relation: NL = 1 2 2 (k − kL ) = wk . Although we have derived these mass formulae by directly analysing the spectrum of elementary string states. (Note that if we had used the nine dimensional canonical metric as deﬁned in equations (4. they can also be derived by analyzing the supersymmetry algebra. (4. we consider states with NR = 0.14) where NL . equivalently. k = 0 or w = 0. as indicated earlier. for deﬁniteness. one does not have the factors of (λ2 ) in the mass formula.14) we see that mass formula for these states takes the form: 2k 2 2k 2 m2 = √ L = √ R · λ2 λ2 (5.15) This is the BPS mass formula for these ultra-short multiplets. NL (NR ) must vanish [64].16) There is no further constraint on w and k.moving sector of the world-sheet respectively18 .1).A. Sen: Duality Symmetries canonical metric. Thus a state with NL = NR = 0 will preserve half of the total number of supersymmetries and will correspond to ultra-short multiplets. and will correspond to short multiplets. .

. It turns out that the total number of short multiplets d(NL ) with NR = 0 for some given value of NL ≥ 1 is given by the formula: d(NL )q NL = NL 1 16 ∞ n=1 1 + qn 1 − qn 8 · (5. − 8 from the NS sector and 8 from the R sector. The overall factor of (1/16) is due to the fact that the lowest level state is only 256-fold degenerate but a single short multiplet requires 16 × 256 states. there is a 16-fold degeneracy of states in each (left. Consider for example states with NR = 0. Thus we get a state with total degeneracy 16 × 16 × 16. it has to act on the Fock vacua with odd fermion i number in order that the states obtained after acting with ψ−1 on the vacua satisfy GSO projection. For example. − 16 from the right moving sector. This gives total degeneracy factor of 8×16 in the left-moving Ramond sector. Thus the net degeneracy of such a state is 16×16 = 256. Due to supersymmetry. NL = 1. showing that there is a unique ultra-short multiplet carrying given charges (kL . For example we get eight states by acting with the transverse bosonic oscillators αi (1 ≤ i ≤ 8). The degeneracy of short multiplets can be found in a similar manner. Let us ﬁrst consider the ultra-short multiplet with k = 0. kR ). and 16 × 16 from the left-moving sector − which is the correct degeneracy of a single short multiplet. for the states with NL = NR = 0. Similar counting can be done for higher values of NL as well. These states have mass m2 = R2 · λ2 (5. Gauge Theory and Strings One can easily calculate the degeneracy of these states by analyzing the spectrum of elementary string states in detail. w = 1. we get an identical factor from the left-moving NS sector as well. In this case there is a 16-fold degeneracy coming from the right-moving sector of the world-sheet. and eight states by acting with the −1 i transverse fermionic oscillators ψ−1 19 .and right-) sector of the world-sheet. There is also an extra degeneracy factor in the left-moving Ramond sector due to the fact that there are many oscillators that can act on the Fock vacuum of the world-sheet theory to give a state at oscillator level NL = 1. There is an 8-fold degeneracy from the Ramond sector Fock vacuum of the leftmoving sector.288 Unity from Duality: Gravity.19) 19 Since ψ i −1 has fermion number one.18) The (1+q n )8 and (1−q n )8 factors in the numerator and the denominator are related respectively to the fact that in the light-cone gauge there are 8 leftmoving fermionic ﬁelds and 8 left-moving bosonic ﬁelds on the world-sheet.

a state carrying p B9µ r units of B9µ charge and r units of B9µ charge. p. (5. an SL(2. r) relatively prime. the w = 1 state will carry one unit of B9µ gauge ﬁeld charge.20) reﬂecting the This converts the w = 1 state.18). B9µ and G9µ charge (momentum along S 1 ) respectively. Thus SL(2. Ref. Z) matrix. r) the degeneracy of such short multiplets. which we shall denote by B9µ p charge carried by the state. c d A similar prediction for the spectrum of BPS states can be made for short multiplets as well. r and k reﬂecting the B9µ . the theory must have a unique ultra-short multiplet with p units of B9µ charge and r units of B9µ charge [61].e. r) are relatively prime. as indicated earlier. cλ + d p r → a c b d p r . p. m2 = R2 |rλ − p|2 .23) . Z) B9µ B9µ → p q r s B9µ B9µ . d(k. The BPS mass formula for these states can be derived by analysing the supersymmetry algebra. For (p. Z) duality predicts that for (p. to a state. λ2 (5.g. r) relatively prime. we see that the SL(2.21) Note that this formula is invariant under the SL(2. p. In this case the state is characterized by three integers p.22) a b is an SL(2. Z) transformation relates these to elementary string states with one unit of winding and k units of momentum along S 1 . Z) duality of type IIB string theory predicts that ∀(p. The condition ps − qr = 1 implies that the pair of integers (p. r)q k = k 1 + qn 1 16 n=1 1 − q n ∞ 8 · (5. [64]). Such states have degeneracy d(k) given in equation (5. Thus in the (8 + 1) dimensional theory obtained by compactifying type IIB on S 1 . Now. r) is independent of p and r and depends on k according to the relation: d(k. r) relatively prime. 1 0 (5. i.A. Sen: Duality Symmetries 289 It is well known that an elementary string acts as a source of the Bµν ﬁeld (see e. Then by following the same logic as before. and is given by. Z) transformation: λ→ where aλ + b . under SL(2. Let us denote by d(k.

partially unbroken supersymmetry. Normally type IIA/IIB string theory contains closed string states only.290 Unity from Duality: Gravity. In type II theory. we must understand the deﬁnition and some of the the properties of these D-branes. in this theory. there is a novel way of constructing a soliton solution that avoids this problem. In particular we need to determine what boundary condition the modes must satisfy at the singularity. and a 2-brane a membrane like object. This construction uses Dirichlet (D-) branes [65. in actual practice. Since open strings satisfy Dirichlet boundary condition in directions transverse to these solitons. these solitons are called D-branes. we shall call a D-brane with Neumann boundary condition in (p + 1) directions (including time) and Dirichlet boundary condition in (9 − p) directions a Dirichlet p-brane. quantizing its zero modes. a 1-brane a string like object. But we can postulate existence of solitonic extended objects in these theories such that in the presence of these solitons. since it can be regarded as a soliton extending along p space-like directions in which we have put Neumann boundary condition. there should be a Hagedorn spectrum of short multiplets p with charge . when one constructs the solution carrying Bµν charge. Thus a state carrying B9µ charge must arise as a soliton. The naive approach will involve constructing such a soliton solution as a solution to the low energy supergravity equations of motion. In order to compute the degeneracy of these solitonic states. 10). (Thus a 0-brane represents a particle like object. Fortunately. there can be open string states whose ends lie on these extended objects (see Fig. Z) symmetry involves explicitly verifying the existence of these states. as identifying the zero modes of a singular solution is not a well deﬁned procedure. with the open string states with ends lying on the soliton corresponding to the (inﬁnite number of) vibrational modes of the soliton. This can in fact be taken to be the deﬁning relation for these solitons. let us consider the following . To see what such a test involves. In particular. However. r A test of SL(2. recall that Bµν arises in the RR sector of string theory. one needs to ensure that the soliton deﬁned this way satisfy all the properties expected of a soliton solution in this theory e.g.) To be more explicit. it turns out to be singular. Of course. existence of static multi-soliton solutions etc. This is the subject to which we now turn. all elementary string states are neutral under RR gauge ﬁelds as can be seen by computing a three point function involving any two elementary string states and an RR sector gauge ﬁeld. and seeing if we recover the correct spectrum of BPS states. Gauge Theory and Strings In other words. Due to this fact it is diﬃcult to proceed further along this line.66].

Z) duality in type IIB string theory: • The Dirichlet p-brane in IIB is invariant under half of the space-time supersymmetry transformations for odd p. xp .24) using various consistency requirements including world-sheet supersymmetry that relates the world-sheet bosons and fermions. Open string states with ends attached to a a) Dirichlet membrane. Let us now summarize some of the important properties of D-branes that will be relevant for understanding the test of SL(2. .A. Note that these boundary conditions break translational invariance along xm . The boundary conditions on the world-sheet fermion ﬁelds are determined from (5. . π) = µ 291 xm 0 for (p + 1) ≤ m ≤ 9 . and extended along x1 . Since we want the full theory to be translationally invariant. Sen: Duality Symmetries boundary condition on the open string: X m (σ = 0. (a) (b) Fig. π) = ∂σ X (σ = 0. We call this a Dirichlet p-brane located at xm = xm 0 (p + 1 ≤ m ≤ 9). 10.24) 0 for where σ denotes the spatial direction on the string world-sheet. To see how this property . b) Dirichlet string. (5. 0 ≤ µ ≤ p. the only possible interpretation of such a boundary condition is that there is a p dimensional extended object situated at xm = xm that is responsible for breaking this transla0 tional invariance. .

Z) duality.25) where Γµ are the ten dimensional gamma matrices.27) • Type IIB (IIA) string theory contains a p-form gauge ﬁeld for even (odd) p. (For p = 5 and 7 these correspond to magnetic dual potentials of Bµν and a respectively. We shall not discuss the details of this computation here. Thus in type IIB string theory Dirichlet p-branes are invariant under half of the space-time supersymmetry transformations for odd p. then a Dirichlet p-brane extending along 1 · · · p direction acts as a source of C01···p . The relevant string world-sheet diagram has been indicated in Figure 11. The open string boundary conditions (5.26) are compatible only for odd p. let us denote by L and R the space-time supersymmetry transformation parameters in type IIB string theory.) This result can be obtained by computing the one point function of the vertex operator for the ﬁeld C in the presence of a D-brane. Γ9 R . More precisely. originating in the leftand the right-moving sector of the world-sheet theory respectively.25) and (5.292 Unity from Duality: Gravity. Γ0 · · · Γ9 R =− R. (5. This means that in type IIB on S 1 (labelled by the coordinate x9 ) a D-string wrapped around the S 1 describes a particle charged under B9µ . if we denote by Cµ1 ···µq the q-form gauge potential. Gauge Theory and Strings arises.24) together with the corresponding boundary conditions on the world-sheet fermions give further restriction on L and R of the form [65]: L = Γp+1 . . Γ0 · · · Γ9 R = R . Γ0 · · · Γ9 L = L. the rank two anti-symmetric tensor ﬁeld Bµν and the rank four anti-symmetric tensor ﬁeld Dµνρσ . (5. From this discussion it follows that a Dirichlet 1-brane (D-string) in type IIB theory carries one unit of charge under the RR 2-form ﬁeld Bµν . in type IIB string theory these p-form gauge ﬁelds correspond to the scalar a. It can be shown that a Dirichlet p-brane carries one unit of charge under the RR (p + 1)-form gauge ﬁeld [65].26) It is easy to see that the two equations (5. As we had seen earlier. (5. L and R satisfy the chirality constraint: Γ0 · · · Γ9 L = L. This then is 0 a candidate soliton carrying charge quantum numbers that is related to 1 1 the state via SL(2.25) is replaced by. . For example. An identical argument shows that in type IIA string theory we have supersymmetric Dirichlet p-branes only for even p since in this theory equation (5. Z) duality 0 . SL(2.

It can be shown that the dynamics of these collective coordinates is described by a (1+1) dimensional supersymmetric quantum ﬁeld theory which . 11. In particular. Thus our task now is as follows: 1 • Quantize the collective coordinates of this soliton. the zero frequency modes (collective modes) of the D-string that are relevant for analyzing the spectrum of BPS states correspond to massless open string states propagating on the D-string. the dynamics of its collective coordinates should be described by a (1 + 1) dimensional ﬁeld theory. • Verify if we get an ultra-short multiplet in this quantum theory. It corresponds to a surface of the topology of a hemisphere with its boundary glued to the D-brane.A. Since the D-string is a one dimensional object. The string world-sheet diagram relevant for computing the coupling of the RR gauge ﬁeld to the D-brane. Sen: Duality Symmetries D-brane 293 X Vertex op. The vertex operator of the RR-ﬁeld is inserted at a point on the hemisphere. By analyzing the spectrum of these open string states one ﬁnds that the collective coordinates in this case correspond to • 8 bosonic ﬁelds y m denoting the location of this string in eight transverse directions. predicts that there should be a unique ultra-short multiplet carrying charge 0 quantum numbers . As we had discussed earlier. all the vibrational modes of the D-string are given by the open string states with ends attached to the D-string. • A U (1) gauge ﬁeld. • 8 Majorana fermions. for RR state String world-sheet Fig.

• 16 fermionic coordinates. is actually a source of B9µ charge! Thus if we restrict to the 0 py = 0 sector then these states carry charge quantum numbers as dis1 cussed earlier. Thus the momentum py conjugate to y is quantized (py = 2πk/a with k integer. • 1 compact bosonic coordinate y. But here since the space direction is compact. but by taking py = 2πk/a. we can actually regard this as a quantum mechanical system instead of a (1 + 1) dimensional quantum ﬁeld theory. A quantum state is labelled by the momenta conjugate to y m (ordinary momenta) and an integer labelling momentum conjugate to y which can be identiﬁed with the quantum number p labelling B9µ charge. we allow it to end on any of the r . p What about states with r > 1? These carry r units of B9µ charge r and hence must arise as a bound state of r D-strings wrapped along S 1 . This establishes p the existence of all the required states of charge predicted by SL(2. which represents electric ﬂux along the D-string. Furthermore. y ≡ A1 dl is a physical variable. the compactness of U (1) makes y to be periodically identiﬁed ((y ≡ y + a) for some a). Thus the ﬁrst question we need to ask is: what is the (1 + 1) dimensional quantum ﬁeld theory governing the dynamics of this system? In order to answer this question we need to study the dynamics of r D-strings.) It can be shown that [62] this momentum. Gauge Theory and Strings is the dimensional reduction of the N = 1 supersymmetric U (1) gauge theory from (9 + 1) to (1 + 1) dimensions. This corresponds to dimensionally reducing the theory to (0 + 1) dimensions. we can ignore all modes carrying momentum along S 1 . This system can be described as easily as a single D-string: instead of allowing open strings to end on a single D-string. Normally in (1 + 1) dimension gauge ﬁelds have no dynamics. The fermionic coordinates satisfy the sixteen dimensional Cliﬀord algebra. It turns out that in looking for ultra-short multiplets. The degrees of freedom of this quantum mechanical system are: • 8 bosonic coordinates y m . Z) 1 symmetry. 1 Due to the compactness of the space direction. Thus quantization of the fermionic coordinates gives 28 = 256 -fold degeneracy. which is precisely the correct degeneracy for a ultra-short multiplet.294 Unity from Duality: Gravity. we can get states carrying charge k quantum numbers as well.

Possible open string states in the presence of three parallel D-strings. Sen: Duality Symmetries 295 Fig. (5. r) which are relatively prime. It turns out that for r coincident D-strings the dynamics of massless strings on the D-string world-sheet is given by the dimensional reduction to (1 + 1) dimension of N = 1 supersymmetric U (r) gauge theory in ten dimensions. precisely as predicted by SL(2.28) where x(i) denotes the location of the i-th D-string. one can show that the probp lem of computing the degeneracy of states reduces to the computation r of certain Witten index in this quantum theory. 1 ≤ i ≤ r. 12. But we can get extra massless states from open strings whose two ends lie on two diﬀerent D-strings when these two D-strings coincide. but also by open strings whose two ends lie on two diﬀerent D-strings. eight scalar ﬁelds and eight Majorana fermions from open strings with both ends lying on that D-string. The situation is illustrated in Figure 12. Thus the dynamics of this system will now be described not only by the open strings starting and ending on the same D-string. For studying the spectrum of BPS states we need to focus our attention on the massless open string states. We shall not go through the details of this analysis. Z) [62]! .A. D-strings situated at xm = xm . or equivalently. for each of the r D-strings we get a U (1) gauge ﬁeld. N = 4 supersymmetric U (r) gauge theory in four dimensions [62]. It turns out that there is a unique ultra-short multiplet for every pair of integers (p. First of all. but just state the ﬁnal result. (i) 2 ≤ m ≤ 9. Following a logic similar to that in the case of a single D-string.

Naturally. for example. k) is done by determining in how many ways the total momentum k can be divided among the various left-moving bosonic and fermionic modes. r) as well. The method of using D-branes to derive the dynamics of collective coordinates has been used to verify the predictions of other duality conjectures involving various string compactiﬁcations. The BPS states come from conﬁgurations where only the left.3 Massless solitons and tensionless strings An interesting aspect of the conjectured duality between the heterotic string theory on T 4 and type IIA string theory on K3 is that at special points in the moduli space the heterotic string theory has enhanced non-abelian gauge symmetry e. At this point the would be massless gauge bosons of the non-abelian gauge theory acquire mass by Higgs mechanism. the duality between heterotic on T 4 and type IIA on K3 [71]. The calculation of the degeneracy d(k. However upon closer examination one realises that this cannot really be a problem [116]. Perturbative type IIA string theory on K3 does not have any such gauge symmetry enhancement. To see this let us consider a point in the moduli space of heterotic string theory on T 4 where the non-abelian gauge symmetry is broken. Z) symmetry of type IIB on S 1 .(or right-) moving modes on S 1 are excited. Instead we need to take into account the modes of the various ﬁelds of the (1 + 1) dimensional ﬁeld theory carrying momentum along the internal S 1 . Gauge Theory and Strings A similar analysis can be carried out for the short multiplets that carry momentum k along S 1 besides carrying the B and B charges p and r [62. we do not have the W ± bosons that are required for enhancing a U (1) gauge group to SU (2). p. the duality between type I and SO(32) heterotic string theory [22]. p. Thus. 5. . since the spectrum of elementary string states does not contain any state charged under the U (1) gauge ﬁelds arising in the RR sector. In order to get these states from the D-brane spectrum. Thus the answer agrees exactly with that predicted by SL(2. we get back the Hagedorn spectrum for d(k. SU (2) at the self-dual radius etc.g. 63]. Among them are self-duality of type II string theory on T 4 [67–70]. At ﬁrst sight this seems to lead to a contradiction. r) of BPS states carrying given charge quantum numbers (p. etc. Z) duality. E8 × E8 or SO(32) in the absence of vacuum expectation value of the internal components of the gauge ﬁelds. r. and appear as BPS states in the abelian theory. except that the elementary string is replaced here by the solitonic D-string. This provides us with a test of the conjectured SL(2. we can no longer dimensionally reduce the (1 + 1) dimensional theory to (0 + 1) dimensions.296 Unity from Duality: Gravity. This counting problem turns out to be identical to the one used to get the Hagedorn spectrum of BPS states in the elementary string spectrum.

For simplicity let us focus on the case of enhanced SU (2) gauge symmetry. These massless D-brane solitons will then provide the states necessary for enhancing the gauge symmetry. we can express this as Ω(A)T Ω(A) for some O(4. after all. Sen: Duality Symmetries 297 As we approach the point of enhanced gauge symmetry. the vector Ω(A) Lα rotates in the twenty four dimensional space.A.29) where α is a 24 dimensional vector belonging to the lattice Λ24 . As we vary M (A) and hence Ω(A) . Since the two tangential directions on the D-2 brane are directed along the two internal directions of K3 tangential to the 2-cycle.29) as m2 = e−Φ (A) /2 T α LΩ(A)T (I24 + L)Ω(A) Lα . For each α we can assign an occupation number n(α) which gives the number of BPS multiplets carrying this speciﬁc set of charges. the images of the W ± bosons in the type IIA theory are given by a D-2 brane wrapped around a certain 2-cycle (topologically non-trivial two dimensional surface) inside K3. the + and the − sign of the charge being obtained from two diﬀerent orientations of the D-2 brane. (I24 + L) has 20 zero eigenvalues. then the masses of these Dbrane states must also vanish as we approach the point in the moduli space where the heterotic theory has enhanced gauge symmetry. since. Although this argument resolves the problem at an abstract level.28). and hence . we do not encounter massless solitons very often in physics. and rewrite equation (5. and represents the U (1) charges carried by this particular state. It can be shown that in the variables deﬁned in Section 4. the masses of these states vanish. 20) matrix Ω(A) . m2 = e−Φ (A) /2 T α (LM (A) L + L)α . 71. First of all. 20) matrix. the vanishing of the masses must be a consequence of the BPS formula. Since the masses of BPS states are determined by the BPS formula. one ﬁnds that at a generic point in the moduli space where SU (2) is broken.3 the BPS formula is given by. Since M (A) is a symmetric O(4. one would like to understand this mechanism directly by analysing the type IIA string theory. 117]. This has been possible through the work of [19. let us examine the BPS formula. To see this more explicitly. (5. this object has no extension in any of the ﬁve non-compact spatial directions. Thus if we are able to ﬁnd the images of these BPS states on the type IIA side as appropriate D-brane states. (5. If for some Ω(A) it is aligned along one of the eigenvectors of (I24 + L) with zero eigenvalue. we shall get massless solitons provided the occupation number n(α) for this speciﬁc α is non-zero.30) As can be seen from equation (4.

It turns out that instead of acquiring enhanced gauge symmetry. vanishes. These arise from taking a D-3 brane of type IIB string theory. since type IIB string theory does not have any D-2 brane solitons which can be wrapped around the collapsed two cycles of K3. obtained by multiplying the tension of the D-2 brane by the area of the two cycle. These are precisely the massless gauge bosons required for the gauge symmetry enhancement in type IIB on K3 × S 1 . type IIB string theory acquires tensionless strings at these special points in the K3 moduli space [119]. This gives us the massless solitons that are required for the gauge symmetry enhancement. Thus from the point of view of the (5 + 1) dimensional theory such a conﬁguration will appear as a string. It turns out that as we approach the point in the moduli space where the theory on the heterotic side develops enhanced SU (2) gauge symmetry. we get tensionless strings. In fact it turns out that there is a one to one correspondence between the enhanced gauge groups. Gauge Theory and Strings behaves like a particle20 . the mass of the wrapped D-2 brane. which are classiﬁed by A-D-E dynkin diagram. which are also classiﬁed by the A-D-E dynkin diagram [19]. Let us compactify this theory on one more circle. A similar mechanism works for getting other gauge groups as well. As a result. and wrapping it on a two cycle of K3. the tension of the string goes to zero. The appearance of enhanced gauge symmetry in type IIA on K3 poses another puzzle. Thus two of the tangential directions of the three brane are directed along the internal directions of K3. The tension of this string is given by the product of the tension (energy per unit three volume) of the three brane and the area of the two cycle on which the three brane is wrapped. But type IIA on K3 × S 1 is T-dual to type IIB on K3 × S 1 . Thus as we approach the singular point on the K3 moduli space where the area of the two cycle vanishes. At this singularity the area of the topologicaly non-trivial 2-cycle mentioned above goes to zero. this theory also has enhanced gauge symmetry when the K3 becomes singular. Does this imply that type IIB on K3 also develops enhanced gauge symmetry at these special points in the K3 moduli space? This does not seem possible. This establishes an explicit physical relationship between A-D-E singularities and A-D-E lie algebras. In other words. . and the singularity type of K3. Upon further compactiﬁcation on a circle we get massless particles from conﬁgurations where this tensionless string is wound around the circle.298 Unity from Duality: Gravity. and the third direction of the three brane is along one of the non-compact spatial directions. thus type IIB on K3 × S 1 must also develop enhanced gauge symmetry when K3 develops singularities. 20 These states were analyzed in detail in [118]. Since such a compactiﬁcation does not destroy gauge symmetry. the K3 on which type IIA theory is compactiﬁed becomes singular.

76]. Quite often G and H do not commute and together generate a much bigger group [23. It turns out that typically these two duality groups do not commute. 78].1 Combining non-perturbative and T -dualities Suppose a string theory A compactiﬁed on a manifold KA has a conjectured duality symmetry group G. Typically part of this self duality group is T -duality. We shall discuss them one by one. The material covered in this section is taken mainly from [75. • It also has a perturbatively veriﬁable T -duality group. but can be “derived” from each other. From the table given in Section 4. We shall illustrate this with a speciﬁc example [23]. Z). Z) that acts non-trivially on the coupling constant. Z) duality symmetry of the ten dimensional type IIB string theory. and in fact generate the full duality symmetry group of type IIB on T n as given in the table of Section 4. Now further compactify this theory on some manifold M. n. In that case. Thus we see that the existence of the full duality symmetry group of type IIB on T n can be infered from the SL(2. the existence of this bigger group of symmetries can be regarded as a consequence of the duality symmetry of A on KA and T -dualities. We have seen that in ten dimensions type IIB string theory has a conjectured duality group SL(2. But quite often the non-perturbative duality transformations in one theory correspond to T -duality in the dual theory .2 Duality of dualities Suppose two theories are conjectured to be dual to each other. and the rest involves non-trivial transformation of the coupling constant.A. 6.5.5 we see that type IIB on T n also has a T -duality group SO(n. We shall now see that many of these conjectures are not independent. Let us call it H. 6. There are several diﬀerent ways in which dualities can be related to each other. 77. and each theory in turn has a conjectured self-duality group. Then the theory A on KA × M is expected to have the following set of duality symmetries: • It inherits the original duality symmetry group G of A on KA . Sen: Duality Symmetries 6 Interrelation between diﬀerent duality conjectures 299 In the last three sections we have seen many diﬀerent duality conjectures and have learned how to test these conjectures. and the perturbatively veriﬁable T -duality symmetries of type IIB on T n . whose existence can be veriﬁed order by order in string perturbation theory.

Let us start with the conjectured duality between heterotic on T 4 and type IIA on K3. and hence can be veriﬁed in this theory order by order in perturbation theory. Again we shall illustrate this with an example [19. Z) that can be veriﬁed using type II perturbation theory. heterotic on T 6 has a T -duality group O(6. Now.22.Z) X SL(2. Z) in heterotic on T 6 and type IIA on K3×T 2.Z) factor of the duality group appears as a T -duality symmetry of the heterotic string theory.Z) TO(6. The full conjectured duality group in both theories is O(6. Thus assuming that T -duality in either theory is a valid symmetry.20. 22.Z) X SL(2. 22. and the duality between the heterotic on T 4 and type IIA on K3.300 Unity from Duality: Gravity. we can establish the existence of the self-duality group O(6. Z) × SL(2. Gauge Theory and Strings DUALITY GROUP FULL HETEROTIC ON T6 O(6. the O(6. Z) × SL(2. Z) factor of the full duality group is a subgroup of the T -duality group in type IIA on K3 × T 2 . Now let us compactify both theories further on a two dimensional torus T 2 .22. how are the T -duality symmetry groups in the two theories embedded in the full conjectured O(6. Z)×SL(2. Z) duality group? This has been illustrated in Figure 13. 24].Z) X SL(2. and vice versa.22. This produces a dual pair of theories: type IIA on K3 × T 2 and heterotic on T 6 . On the other hand.Z) Fig. In particular we ﬁnd that the SL(2. 13. the full self duality group in both theories follows from the conjectured duality between the two theories.Z) O(4. Z)×SL(2. . As a result. Now the question we would like to address is. Z)×SL(2. type IIA on K3×T 2 has a T -duality group O(4. 22.Z) TYPE II ON K3 X T2 O(6. On the other hand . 20. Z). 22.Z) X SL(2. The embedding of the T -duality groups in the full duality group in heterotic on T 6 and type IIA on K3 × T 2 . Z) that can be veriﬁed using heterotic perturbation theory.22. and hence can be veriﬁed order by order in perturbation theory in this theory.

Now construct a pair of new manifolds MA . KB are two diﬀerent manifolds (in general). the independent ones are: 1. Sen: Duality Symmetries 301 Using the results of this and the previous subsection. 6. KB vary slowly over B and are related to each other via the duality map that relates (A on KA ) to (B on KB ). The moduli of KA . This situation has been illustrated in Figure 14. and 3. SL(2.3 Fiberwise duality transformation A on KA B on KB B A on MA B B on M B Fig. Then we would expect a duality Theory A on MA ↔ Theory B on MB by applying the duality transformation ﬁberwise. Here A and B are two of the ﬁve diﬀerent string theories in D = 10. . type I ↔ SO(32) heterotic in D = 10. 14. the two theories are equivalent due to the equivalence of the theories living on the ﬁber (× any manifold). In this subsection we shall describe the idea of constructing dual pairs of theories using ﬁberwise duality transformation [77]. we see that so far among all the conjectured non-perturbative duality symmetries. This duality involves a precise map between the moduli spaces of the two theories. KB . In each local neighbourhood of the base manifold B. Application of ﬁberwise duality transformation. Z) of type IIB in D = 10. and KA .A. Suppose (Theory A on KA ) has been conjectured to be dual to (Theory B on KB ). and erecting at every point on B a copy of KA . This then gives rise to a new duality conjecture. MB by starting from some other manifold B. 2. We shall now show how to “derive” 3) from 1) and 2). IIA on K3 ↔ heterotic on T 4 . Thus we would expect the theories A on MA and B on MB to be equivalent.

then there is no identiﬁcation of the points in the copy of KA that is sitting at Q. Q ∈ D) to (hA (p). and compare the two quotient theories (A on KA × D/hA · g) and (B on KB × D/hB · g) (KA ×D/hA ·g) is obtained from the product manifold KA ×D by identifying points that are related by the Z2 transformation hA · g. we can derive the existence of many new duality symmetries from a given duality symmetry. note that since hA · g takes a point (p ∈ KA . this appears to be the case in all known examples. This shows that the ﬁber is KA and not KA /hA . assuming that this is the case. (KA × D/hA · g) admits a ﬁbration with base D/g and ﬁber KA . Suppose we have a dual pair (A on KA ) ↔ (B on KB ). Representation of a Z2 orbifold as a ﬁbered space. Thus we can now apply ﬁberwise duality transformation to derive a new . A special case of this construction involves Z2 orbifolds. if we focus our attention on a deﬁnite point Q on D.) Is the duality between (A on MA ) and (B on MB ) still valid? We might expect that even in this case the duality between the two theories holds since the singularities occur on subspaces of “measure zero”. which is a closed cycle on D/g. Gauge Theory and Strings KA Q KA p g(Q) Z2 hA (p) KA Q=g(Q) D D/g Fig. As we go from Q to g(Q). 15. hA (p)). The second theory. In particular. Conversely. Although there is no rigorous argument as to why this should be so. g(Q)).302 Unity from Duality: Gravity. As shown in this ﬁgure. Now compactify both theories on another manifold D with a Z2 isometry generated by g. p) on D × KA to the point (g(Q). the ﬁber gets twisted by the transformation hA . Now suppose that at some isolated points (or subspaces of codimension ≥1) on B the ﬁbers KA and KB degenerate. The Z2 transformation relates the point (Q. hA and hB are mapped to each other under duality. (We shall see some explicit examples of this later. B on (KB × D)/(hB · g) has an identical structure. Further suppose that (A on KA ) has a Z2 symmetry generated by hA . This situation is illustrated in Figure 15. Then the dual theory must also have a Z2 symmetry generated by hB .

but as has been argued before. This gives the duality: (IIB on T 4 /(−1)FL · I4 ) ↔ (IIB on T 4 /Ω · I4 ). The action of S. We shall now illustrate this construction in the context of a speciﬁc example. Let S denote the SL(2. hB to be Ω. This has a conjectured SL(2. However. hA to be (−1)FL .2) We are now ready to apply our formalism. This action turns out to be identical to that of Ω. D to be T 4 . Similarly in KB × D/hB · g there is an identiﬁcation of points (p . (−1)FL and Ω on the massless bosonic ﬁelds in this theory were described in Section 4. Let us make R → (1/R) duality transformation on one of the circles of T 4 in the theory on the left hand side. since the action of S on the massive ﬁelds is not known. Thus at P0 the ﬁbers degenerate to KA /hA and KB /hB respectively.4. P0 ).A. (6. Sen: Duality Symmetries duality: (A on KA × D/hA · g) ↔ (B on KB × D/hB · g) 303 Note that if P0 ∈ D is a ﬁxed point of g (i. as we have discuused earlier. A similar result holds for their action on the massless fermionic ﬁelds as well. We shall now bring this duality into a more familiar form via T -duality transformation. (6. one can deﬁne this action in such a way that the actions of S(−1)FL S −1 and Ω are identical on all states. P0 ) and (hA (p). P0 ) and (hB (p ). we still expect the duality to hold since these are points of measure zero on T 4 . At these points the argument in support of duality between the two theories breaks down. and g to be the transformation I4 that changes the sign of all the coordinates on T 4 . We start with type IIB string theory in ten dimensions. we would expect that the two quotient theories are still dual to each other [78]. From this one can explicitly compute the action of S(−1)FL S −1 on these massless ﬁelds. but this does not prevent us from applying our method of constructing dual pairs. (B on KB ) to be type IIB in D = 10 transformed by S. We take (A on KA ) to be type IIB in D = 10.1) Recall that this theory also has two global discrete symmetries (−1)FL and Ω. Finally. This gives: S(−1)FL S −1 = Ω . Z) element 0 1 −1 0 . P0 ). if g(P0 ) = P0 ) then in KA × D/hA ·g there is an identiﬁcation of points (p.e. Also there are sixteen ﬁxed points on T 4 under I4 where the application of ﬁberwise duality transformation breaks down. Note that in this case the ﬁbers KA and KB are points. This converts type IIB . since these are points of “measure zero” on D. Z) symmetry.

Suppose a string theory compactiﬁed on a manifold M1 × M2 has a self-duality symmetry group G. But by (heterotic – type I) duality in ten dimensions this is dual to heterotic on T 4 . Z) duality symmetry of type IIB string theory from the conjectured duality between type I and SO(32) heterotic string theories. There are many other applications of ﬁberwise duality transformation. Thus the theory on the right is T -dual to type IIB on T 4 /Ω. the argument can be generalized to establish this duality at a generic point in the moduli space as well [78]. Now consider the limit when the size of M2 goes to inﬁnity.4 Recovering higher dimensional dualities from lower dimensional ones So far we have discussed methods of deriving dualities involving compactiﬁed string theories by starting with the duality symmetries of string theories in higher dimensions. which can again be seen by studying the action of these transformations on the massless ﬁelds.304 Unity from Duality: Gravity. However. Let us now take the theory on the right hand side and make R → (1/R) duality transformation on all four circles. A priori this procedure does not appear to be very useful. which can be checked by explicitly studying the action of these transformations on the various massless ﬁelds.e. acting on a conﬁguration where M2 is big. Thus we have “derived” the duality (Type IIA on K3) ↔ (Heterotic on T 4 ) from other conjectured dualities in D = 10. 6. This takes type IIB theory to type IIB theory. since one normally likes to derive more complicated duality transformations of lower dimensional theories from the simpler ones in the higher dimensional theory. Gauge Theory and Strings theory to type IIA. and . This of course is just a special case of type IIA on K3. Thus the theory on the left hand side is T -dual to type IIA on T 4 /I4 . Some of them will be discussed later in this review. gives us back a conﬁguration where M2 is big. But we can also proceed in the reverse direction. Thus we would expect that H is the duality symmetry group of the theory in the decompactiﬁcation limit. there may be a subgroup H of G that commutes with this limit. A generic element of G. Since type I string theory can be regarded as type IIB string theory modded out by Ω. any element of this subgroup.e. But this transforms Ω · I4 to Ω. This also transforms (−1)FL · I4 to I4 . i. Although this way the duality has been established only at a particular point in the moduli space (the orbifold limit of K3). will convert this conﬁguration to one where M2 has small or ﬁnite size. The same argument can be extended to the case of a pair of dual theories. of the original string theory compactiﬁed on M1 . acting on this conﬁguration. i. But we shall now show how this procedure can be used to derive the SL(2. we see that the theory on the right hand side is type I on T 4 .

for details. 7 Duality in theories with less than sixteen supersymmetry generators So far our discussion has been focussed on theories with 16 or more supersymmetry charges. and the other one is associated with the R → (1/R) duality symmetries on the two circles. this must also be a symmetry of type I on T 2 . Now take the limit where the size of the tetrahedron goes to inﬁnity. Z) factors is associated with the global diﬀeomorphism of T 2 . we conclude that SL(2. Z) × SL(2.4.) Geometrically. In the next section we shall see more examples of dualities which can be derived from the ones that we have already discussed. with an added twist of (−1)FL · Ω as we go around any of the four vertices of the tetrahedron (the ﬁxed points of I2 ). where I2 denotes the reversal of orientation of both the circles of T 2 . − the one between type I and SO(32) heterotic string theories in ten dimensions. Z) is a subgroup of the self-duality group of type IIB on T 2 /Ω.A. Z) factors commute with this limit. Z) subgroup of this T -duality group. this model describes type IIB string theory compactiﬁed on the surface of a tetrahedron (which is geometrically T 2 /I2 ). Since this limit gives us back the decompactiﬁed type IIB string theory. see [75]. Z) groups becomes part of the diﬀeomorphism group of type IIB string theory and does not correspond to anything new. (This can be seen by studying the action of various transformations on the massless ﬁelds. we conclude that type IIB string theory in ten dimensions has a self-duality group SL(2. Sen: Duality Symmetries 305 T -duality symmetries of the heterotic string theory. We start with the duality between type I on T 2 and heterotic on T 2 that follows from the duality between these theories in ten dimensions. Z) factor represents the S-duality transformation discussed in Section 4.4. but the other SL(2. Thus we see that all the dualities discussed so far can be “derived” from a single duality conjecture. By the “duality of dualities” argument. Z) × SL(2.18. Let us now make an R → (1/R) duality transformation on both the circles of this T 2 . It turns out that both the SL(2. One of these two SL(2. and Ω to (−1)FL · Ω · I2 . Now heterotic string theory on T 2 has a T -duality group O(2. As was pointed out in Section 4.Z). Since type I string theory can be regarded as type IIB string theory modded out by the world-sheet parity transformation Ω discussed in Section 4. Z) is a subgroup of the self-duality group of type IIB on a tetrahedron. This converts type IIB on T 2 to type IIB theory compactiﬁed on a dual T 2 . for these theories the non-renormalization theorems for the low energy eﬀective action and . Z) × SL(2. One of these SL(2. We shall focus our attention on an SL(2. Thus we conclude that SL(2. Z). We shall describe the main steps in this argument.

Type IIA/IIB on Calabi-Yau 3-folds: in our convention an n-fold describes an n complex or 2n real dimensional manifold. Thus we have a theory with eight supersymmetry generators. 7. 108].306 Unity from Duality: Gravity. In ten dimensions type II theories have 32 supersymmetry generators. There are several ways to get theories with N = 2 supersymmetry in four dimensions. Gauge Theory and Strings the spectrum of BPS states are particularly powerful. This makes it easy to test duality conjectures involving these theories. giving rise to N = 2 supersymmetry. Thus we are left with 8 supersymmetry generators in D = 4. In this section we shall extend our discussion to theories with eight supercharges. Compactiﬁcation on a Calabi-Yau 3-fold breaks 3/4 of the supersymmetry.1 Construction of a dual pair of theories with eight supercharges For deﬁniteness we shall focus our attention on N = 2 supersymmetric theories in four dimensions. 2. Compactiﬁcation on K3×T 2 breaks half of the supersymmetry. [77. However we shall begin by describing the systematic approach. It is also possible to construct more general class of four dimensional heterotic string theories with the same number of supersymmetries where the background does not have the product structure K3 × T 2 [79. 79. The steps involved in this construction are as follows: • Start from the conjectured duality (Type IIA on K3) ↔ (Heterotic on T 4 ). they are still powerful enough to provide us with some of the most striking tests of duality conjectures involving these theories. The systematic construction of such dual pairs can be carried out by application of ﬁberwise duality transformation as described in the last section. The question we would like to ask is: is it possible to construct pairs of N = 2 supersymmetric type II and heterotic string compactiﬁcations in four dimensions which will be non-perturbatively dual to each other? Historically such dual pairs were ﬁrst constructed by trial and error [79] and then a more systematic approach was developed [77. and then describe how one tests these dualities. We shall see that these theories have a very rich structure. Examples of such theories are provided by N = 2 supersymmetric theories in four dimensions. again giving N = 2 supersymmetry in four dimensions. The material covered in this section is based mainly on refs. Two of them are: 1. Heterotic string theory on K3 × T 2: in ten dimensions heterotic string theory has sixteen supersymmetry generators. and although the non-renormalization theorems are less powerful here. 80–83]. 80]. .

Construction of dual pair of N = 2 supersymmetric string theories in four dimensions from the dual pair of theories in six dimensions. • Choose a CP 1 base. Thus we get a duality map (Type IIA on CY) ↔ (Heterotic on K3×T 2). knowing how K3 varies over CP 1 . This model is expected to be dual to the type IIA string theory on the Calabi-Yau manifold that we started with. This gives heterotic string theory on a manifold obtained by varying T 4 on CP 1 according to the duality map. • For type IIA on each such Calabi-Yau 3-fold we can get a dual heterotic compactiﬁcation by replacing the type IIA on K3 by heterotic on T 4 on each ﬁber according to the duality map. we can ﬁnd out how on the heterotic side the background gauge ﬁelds on T 4 vary as we move along CP 1 .A. This gives the gauge ﬁeld conﬁguration on K3 × T 2 . One can construct a whole class of Calabi-Yau manifolds this way by choosing diﬀerent ways of varying K3 over CP 1 . . Diﬀerent Calabi-Yau manifolds will give rise to diﬀerent gauge ﬁelds on K3 × T 2 . 16. On the heterotic side the moduli involve background gauge ﬁelds on T 4 besides the shape and size of T 4 . Note that the original duality map gives a precise relationship between the moduli of type IIA on K3 and heterotic on T 4 . Typically this manifold turns out to be K3 × T 2 or some variant of this. Thus for a speciﬁc Calabi-Yau. • Construct a Calabi-Yau 3-fold by ﬁbering K3 over the base CP 1 . Sen: Duality Symmetries 6D Duality map K3 T 4 307 CP 1 CP 1 IIA on CY Heterotic on K3 X T 2 Fig. This construction has been illustrated in Figure 16.

has known action on the harmonic forms ω (p) . It turns out that the action on the heterotic side is given by: • exchanging the gauge ﬁelds in the two E8 factors. Now. (p) . K3×T 2/(hA ·g) can be shown to describe a Calabi-Yau manifold. Furthermore. This translates into a similar action on the ﬁelds Aµ deﬁned in equation (4. hA leaves the other two gauge ﬁelds. hA . As discussed there. we need to determine hB . B9µ ) with (G8µ . For this particular example. We can now translate this into an action on the gauge ﬁelds in heterotic on T 4 . • changing the sign of (G7µ and B7µ ).3. • exchanging (G9µ . In order to determine hB . coming from the ten dimensional gauge ﬁeld Aµ and the dual of Cµνρ invariant. Gauge Theory and Strings We shall illustrate this procedure with the example of a pair of Z2 orbifolds of the form [80]: (IIA on K3 × T 2 /hA · g) ↔ (Heterotic on T 4 × T 2 /hB · g) where g acts on T 2 by changing the sign of both its coordinates. By our previous argument relating orbifolds to ﬁbered spaces. Thus the theory on the left-hand side corresponds to type IIA string theory compactiﬁed on this Calabi-Yau manifold. being a geometric transformation on K3. and hB is the image of this transformation on the heterotic side. B8µ ). we need to study the relationship between the ﬁelds appearing in type IIA on K3 and heterotic on T 4 . We shall focus our attention on the gauge ﬁelds.308 Unity from Duality: Gravity.36). and. hA corresponds to • exchanging ten of the ω (p) with ten others and • changing the sign of two more ω (p) . hA is a speciﬁc involution of K3 known as the Enriques involution. these two theories are expected to be dual to each other via ﬁberwise duality transformation. 22 of the gauge ﬁelds come from decomposing the three form ﬁeld along the harmonic two forms on K3. We shall now describe this procedure in some detail. The low energy eﬀective action of both the theories and the origin of the various massless ﬁelds in these theories were discussed in Section 4. in the type IIA on K3. In order to determine the theory on the heterotic side.

as they undergo quantum corrections in general. • x7 → −x7 . Matter multiplets in N = 2 supersymmetric theories in four dimensions are of two types. Quite often using mirror symmetry [109] we can also relate this to IIB string theory on a mirror Calabi-Yau manifold. one complex scalar. . Furthermore. and 21 Here we are regarding this theory as the E × E heterotic string theory compactiﬁed 8 8 on T 4 . Unlike in the case of theories with sixteen supercharges. By the duality between the two heterotic string theories upon compactiﬁcation on a circle. 22 We need to add to this a non-geometrical shift involving half of a lattice vector in Λ24 in order to get a modular invariant theory on the heterotic side.A. one cannot directly compare the tree level low energy eﬀective action in the two theories. This is hB 22 . It turns out that modding out heterotic string theory on T 6 by the transformation hB · g produces an N = 2 supersymmetric theory. the next question will be: how do we test if these theories are really dual to each other? After all. • x8 ↔ x9 . (For a review. 7.) They are • vector multiplet containing one vector. Using the idea of ﬁberwise duality transformation we can construct many more examples of heterotic – type IIA dual pairs in four dimensions with N = 2 supersymmetry. see [88]. and two Majorana fermions. this is equivalent to SO(32) heterotic string theory compactiﬁed on T 4 . as we shall now describe. there is no rigorous proof that ﬁberwise duality transformation always produces a correct dual pair of theories. This transformation is not visible in perturbative type IIA theory. Thus this construction gives a type II – heterotic dual pair with N = 2 supersymmetry. as we have seen. Hence the spectrum computed at weak coupling cannot always be trusted at strong coupling. in this theory the spectrum of BPS saturated states can change discontinuously as we move in the moduli space [88]. Nevertheless there are some non-renormalization theorems which allow us to test these proposed dualities. Sen: Duality Symmetries 309 This translates into the following geometric action in heterotic string theory on T 4 :21 • exchange of two E8 factors in the gauge group.2 Test of duality conjectures involving theories with eight supercharges Given such a dual pair of theories constructed by application of ﬁberwise duality transformation. particularly when the ﬁber degenerates at some points/regions in the base.

Let us consider a theory at a generic point in the moduli space where the massless matter ﬁelds include only abelian gauge ﬁelds and neutral hypermultiplets. Thus the hypermultiplet 23 There are however strong restrictions on what kind of metric G V and GH should describe. the dilaton is in the vector multiplet. Gauge Theory and Strings • hypermultiplet containing two complex scalars and two Majorana fermions. But in string theory we have some extra ingredient [77. Recall that the coupling constant in string theory involves the dilaton.1) where GV and GH are appropriate metrics in the vector and the hypermultiplet moduli spaces. The N = 2 supersymmetry requires that there is no coupling between the vector and the hypermultiplets in those terms in the low energy eﬀective action Seﬀ which contain at most two space-time derivatives [84]. On the other hand in heterotic on K3 × T 2 . . Now consider the following two special cases. is exact. m¯ α¯ (7.310 Unity from Duality: Gravity. these restrictions do not ﬁx GV and GH completely. 2. In particular GV must describe a special Kahler geometry [84. However.and the vector. since each term may be independently modiﬁed by quantum corrections23 . Thus quantum corrections to a given term must involve a coupling to the dilaton. In type IIA/IIB string theory on Calabi-Yau manifold the dilaton belongs to a hypermultiplet. 1. calculated at the tree level. and ψ denote the complex scalars in the hypermultiplet. The kinetic terms of the vectors and the fermionic ﬁelds are related to these scalar kinetic terms by the requirement of N = 2 supersymmetry. 79]. Thus the scalar kinetic terms appearing in the Lagrangian density associated with Seﬀ must be of the form: ¯ ¯ GV n (φ)∂µ φm ∂ µ φn + GHβ (ψ)∂µ ψ α ∂ µ ψ β . Let φ denote the complex scalars in the vector multiplet. The dilaton belongs to a hypermultiplet. Then there can be no correction to the vector multiplet kinetic term since such corrections will give a coupling between the dilaton and the vector multiplet.multiplet moduli spaces by itself is not of much help. whereas GH must describe a quaternionic geometry [111]. Thus in these theories the vector multiplet kinetic term. The dilaton belongs to a vector multiplet. 110]. This decoupling between the hyper. In this case the same argument shows that there can be no correction to the hypermultiplet kinetic term.

Indeed. we can eﬀectively get an N = 4 supersymmetric Yang-Mills theory. Using this information we can adopt the following strategy for testing duality24 . calculated explicitly in the heterotic string theory. . with the expansion of the exact answer. From the exact answer calculated from the type II side we know what this contribution should be. 25 This is in the same spirit as in the case of toroidal compactiﬁcation of heterotic string theory.85. This can be carried out using the method developed by 24 Here we describe the test using the vector multiplet kinetic term.86]! This agreement is quite remarkable. Typically the expansion involves tree. Sen: Duality Symmetries 311 kinetic term. the agreement between the two answers is a consequence of highly non-trivial mathematical identities. Take a type II – heterotic dual pair and calculate the vector multiplet kinetic term exactly from the tree level analysis on the type II side. and non-perturbative terms.) Thus one can compare the expected tree and one loop terms. one loop. Using the map between the ﬁelds in the type II and the heterotic theory. since there is no non-perturbative formulation of string theory. 2. However. Given that the tree and one loop results in the heterotic string theory agree with the expansion of the exact result on the type II side. but a similar analysis should be possible with the hypermultiplet kinetic term as well. But we cannot calculate it directly on the heterotic side. since the one loop calculation is highly non-trivial on the heterotic side. calculated at the tree level. Thus now the calculation of these non-perturbative eﬀects on the heterotic side reduces to a calculation in the N = 2 supersymmetric ﬁeld theory. one can take an appropriate limit in which the stringy eﬀects on the heterotic side disappear and the theory reduces to some appropriate N = 2 supersymmetric quantum ﬁeld theory25 . one might ask if a similar agreement can be found for the non-perturbative contribution from the heterotic string theory as well. 1. we can rewrite the exact vector multiplet kinetic term in terms of the heterotic variables. by going near a special point in the moduli space. The results of the above calculation in heterotic and type II string theories agree in all the cases tested [79. 3. is exact. In particular the heterotic variables include the heterotic dilaton ΦH which is in the vector multiplet. (There is no perturbative contribution in the heterotic theory beyond one loop due to some Adler-Bardeen type non-renormalization theorems. and involves integrals over the moduli space of the torus. So we can now expand the exact answer in powers of eΦH and compare this answer with the explicit calculations in heterotic string perturbation theory. where.A.

− (12)4 (8. Take type IIA string theory in ten dimensions. It is well known that this can be obtained from the dimensional reduction of N = 1 supergravity in eleven dimensions [90]. 89].10 ) measured in the (S) supergravity metric gMN and ignore (for the time being) the Kaluza-Klein modes carrying momentum in the internal direction. The bosonic ﬁelds in (S) N = 1 supergravity theory in eleven dimensions consist of the metric gMN (S) and a rank three anti-symmetric tensor ﬁeld CMNP (0 ≤ M. the relationship between the two theories is as follows. N ≤ 10). More speciﬁcally. Gauge Theory and Strings Seiberg and Witten [88]. In writing down the above equation we have set the eleven dimensional Planck mass to unity (or equivalently we can say that we have absorbed it into a redeﬁnition of the metric. This theory is now known as M-theory.1) where G(S) ∼ dC (S) is the four form ﬁeld strength associated with the three form ﬁeld C (S) . Besides providing a non-trivial test of string duality. One such theory is a conjectured theory living in eleven dimensions. However. 8. this also shows that the complete Seiberg-Witten [88] results (and more) are contained in the classical geometry of Calabi-Yau spaces! 8 M-theory So far we have discussed dualities that relate known string theories. In this section we shall give a brief description of this theory following references [19. The low energy eﬀective action of this theory is non-chiral N = 2 supergravity in ten dimensions.1 M-theory in eleven dimensions The arguments leading to the existence of M-theory goes as follows [19.312 Unity from Duality: Gravity. Then the eﬀective action in the dimensionally reduced theory agrees with that of type IIA . sometime analysis similar to those that lead to various duality conjectures can also lead to the discovery of new theories. The bosonic part of the action of this theory is given by [112] SSG = 1 1 −g (S) R(S) − G(S)2 d11 x 8 (2π) 48 1 µ0 ···µ10 (S) (S) (S) ε Cµ0 µ1 µ2 Gµ3 ···µ8 Gµ7 ···µ10 .) Let us now compactify this supergravity theory on a circle of radius R(∼ (S) g10. Again there is perfect agreement with the results from the type II side [87]. 93–98].

One might ask if there are more precise tests involving the spectrum of BPS states. as discussed so far. Part of the conjecture is just the deﬁnition of M-theory as the strong coupling limit of type IIA string theory.10 = eΦ/3 . (S) Cµνρ (S) (S) gµν 313 e−Φ/12 gµν Bµν .33) under the identiﬁcation [90]: g10. and can be shown to belong to the 256 dimensional ultra-short (S) . The evidence for the existence of an eleven dimensional theory. This corresponds to strong coupling limit of the type IIA string theory. has been analogous to the evidence for various duality conjectures based on the comparison of their low energy eﬀective action. We are using the convention that Φ = 0 corresponds to compactiﬁcation on a circle of unit radius. This leads one to the conjecture [19.10 approaches ∞. The relationship between M-theory and various other supergravity/string theories. C10µν (S) (0 ≤ µ. The non-trivial part of the conjecture is that it describes a Lorentz invariant theory in eleven dimensions. These are BPS states. 17. This theory has been called M-theory.A. Φ → ∞. Here denotes equality up to additive terms involving second and higher 1 S compactification M-theory Type IIA string theory Low energy limit D=11 supergravity S Type IIA supergravity 1 compactification + low energy limit Low energy limit Fig. g10µ (S) e2Φ/3 Aµ . Sen: Duality Symmetries string theory given in (4. (8. powers in ﬁelds. Note that as the radius R ∼ g10. whose low energy limit is 11-dimensional N = 1 supergravity. There are indeed such tests.2) Cµνρ . 89] that in the strong coupling limit type IIA string theory approaches an 11 dimensional Lorentz invariant theory. M theory on S 1 will have Kaluza-Klein modes representing states in the eleven dimensional N = 1 supergravity multiplet carrying momentum along the compact x10 direction. ν ≤ 9) . The situation is illustrated in Figure 17.

IIA duality. As usual. The main obstacle to this analysis is that a charge k state has the same energy as k charge 1 states at rest.314 Unity from Duality: Gravity. Dynamics of collective coordinates of k D0 branes is given by the dimensional reduction of N = 1 supersymmetric U (k) gauge theory from (9 + 1) to (0 + 1) dimensions. Thus one should be able to ﬁnd these states in type IIA string theory on . Thus for every integer k we should ﬁnd such BPS states in type IIA string theory in (S) ten dimensions. Gauge Theory and Strings representation of the supersymmetry algebra. k2 ) relatively prime. is k1 R1 2 (S) + k2 R2 2 · (8. For k = 2 such a bound state with the correct degeneracy has been found [91]. such a state has strictly less energy than the sum of the masses of any other set of states with the same total charge [92]. Such states are diﬃcult to study. we see that these states cannot come from elementary string states. as elementary string excitations are neutral under RR sector gauge ﬁelds. The analysis can be simpliﬁed by compactifying M-theory on T 2 and considering the Kaluza-Klein modes carrying (k1 . Thus the bound states we are looking for sit at the bottom of a continuum. Finding these bound states is much more diﬃcult than the bound state problems discussed earlier. up to a proportionality factor. This theory has sixteen fermion zero modes whose quantization leads to a 28 = 256 fold degenerate state. However Dirichlet 0-branes in this theory do carry Aµ charge. Thus we see that we indeed have an ultra-short multiplet with unit Aµ charge. What about states with k > 1? In type IIA string theory these must arise as bound states of k D0-branes. Since g10µ gets mapped to Aµ under the M-theory . Thus the number of ultra-short multiplets with k-units of Aµ charge is determined in terms of the number of normalizable supersymmetric ground states of this quantum mechanical system. these states must carry k units of Aµ charge in type IIA string theory. the collective coordinate dynamics of the 0-branes is determined from the dynamics of massless open string states with ends lying on the D0-brane. In M-theory these states carry k units of g10µ charge. The charge quantum number characterizing such a state is the momentum (k/R) along S 1 . the mass of such a state. and have radii R1 and R2 respectively. If we now recall that in type IIA string theory Aµ arises in the RR sector. Assuming that the two S 1 ’s are orthogonal. as predicted by the M-theory . k2 ) units of momenta along the two S 1 ’s. and in this case is described by the dimensional reduction of N = 1 super-Maxwell theory from (9 + 1) to (0 + 1) dimensions.IIA duality conjecture. The analysis for higher k still remains to be done. In particular the state with k = 1 corresponds to a single Dirichlet zero brane.3) For (k1 .

this would mean that we should have Lorentz invariance in coordinates x0 . x10 . and the action of SL(2. Z) strong-weak coupling duality inherited from ten dimensional type IIB string theory. But in M-theory on T 2 it is simply a consequence of the diﬀeomorphism invariance of the 11dimensional theory. . . . Z) in type IIB string theory. • Then from the conjectured SL(2.A. deﬁned as the strong coupling limit of IIA. But we know that IIA on S 1 is related by T duality to IIB on S 1 . x9 are non-compact.2. In the limit when the radius of both the compact circles are taken to be large. according to the conjecture. . x9 when all the coordinates x0 . Z) symmetry in M-theory is simply the group of global diﬀeomorphisms of T 2 [61. . (For a review of 26 In fact. . Turning this analysis around we see that this also supports the ansatz that M-theory. these states can be shown to be in one to one correspondence to a class of supersymmetric vacua in a (1 + 1) dimensional supersymmetric gauge theory compactiﬁed on a circle26 . The SL(2. According to M-theory – type IIA duality. Consider M-theory on T 2 .2 Compactiﬁcation of M-theory Given the existence of M-theory. . . is equivalent to M-theory on T 2 ) without encountering the diﬃculties mentioned earlier. There are also other consistency checks on the proposed M-theory – IIA duality. is a fully Lorentz invariant theory in eleven dimension. Thus we have a duality between M-theory on T 2 and IIB on S 1 . it is dual to IIA on S 1 . All such states have been found with degeneracy as predicted by the M-theory – IIA duality. we can now construct new vacua of the theory by compactifying M-theory on various manifolds. Z) duality symmetry of type IIB string theory we know that we have an exchange symmetry between the 9th and the 10th coordinate of M-theory when these coordinates are compact. 8. Thus we again have an example of “duality of dualities”. . It turns out that this SL(2. 94]. from Lorentz invariance of type IIA string theory we know that we have Lorentz invariance in coordinates x0 . 93. Thus one might ask. Z) of IIB is a non-perturbative symmetry. By following the same kind of argument. Now IIB on S 1 has an SL(2. what does it correspond to in M-theory on T 2 ? One can ﬁnd the answer by using the known map between the massless ﬁelds in the two theories. Sen: Duality Symmetries 315 S 1 (which. these states are related via an R → (1/R) duality transformation to the ultra-short multiplets in type IIB on S 1 discussed in Section 5. . The argument goes as follows: • First of all.

The supergravity sector. is expected to be responsible for supersymmetry breaking. The essential feature of this strong coupling limit is the emergence of Lorentz invariance in one higher dimension.) For example. known as the hidden sector or the shadow world.316 Unity from Duality: Gravity. The duality between M-theory on S 1 /Z2 and the E8 × E8 heterotic string theory is particularly amusing. we can consider M-theory compactiﬁed on K3. Now in the conventional heterotic string compactiﬁcation on Calabi-Yau spaces. Gauge Theory and Strings compactiﬁcation of eleven dimensional supergravity. Calabi-Yau. sits in the bulk. For example. for (S) odd n this is also accompanied by a reversal of sign of CMNP . These can all be regarded as appropriate strong coupling limits of type IIA compactiﬁcation on the same manifold. In the M-theory picture these two sectors are physically separated in space. The second E8 . all the observed gauge bosons and charged particles come from one E8 and are neutral under the second E8 [128]. It turns out that the two E8 gauge multiplets arise from “twisted sector” of the theory and sit at the two ends of this line segment. see [138]. upon further compactiﬁcation on a circle. Here the Z2 transformation acts by reversing the orientation of S 1 . Each of these duality conjectures satisfy the consistency condition that the theory on the right hand side. Such a theory cannot be constructed by conventional compactiﬁcation of type IIA string theory at weak coupling. Of course in many cases these non-perturbative vacua are related to perturbative string vacua by conjectured duality relations. In other words. S 1 /Z2 denotes a real line segment bounded by the two ﬁxed points on S 1 . These duality conjectures can be arrived at by using arguments very similar to those used in arriving at string duality conjectures. and various orbifolds. Some examples of such conjectured dualities are given below [95–98]: M-theory on ↔ (E8 × E8 ) heterotic in D = 10 S 1 /Z2 K3 T /Z2 5 ↔ Heterotic/Type I on T 3 ↔ IIB on K3 ↔ Type I/Heterotic on T 7 ↔ Type IIB on T 8 /Z2 T 8 /Z2 T 9 /Z2 In each case Z2 acts by reversing the sign of all the coordinates of T n . together with a change of sign of the three (S) form ﬁeld CMNP . the real world and the shadow world live at two ends of the line and interact only via the exchange of supergravity multiplets . M-theory on a Calabi-Yau manifold gives a ﬁve dimensional theory with N = 1 supersymmetry [137]. But in general these cannot be regarded as perturbative string vacua. on the other hand. is dual to type IIA string theory compactiﬁed on the manifold on the left hand side.

since if this dimension is too large. . . . and σ denotes the transformation CMNP → −CMNP . B on KB to be type IIA string theory. hA to be I1 · σ. . if we take the theory on the right hand side and make an R → (1/R) duality transformation on one of the circles. which is a special case of type IIB on K3. x9 ). The theory on the left hand side is M-theory on T 5 /Z2 . . On the other hand. it converts • type IIA theory to type IIB theory. The Z2 generator is I5 · σ where I5 changes the sign of all ﬁve coordinates (S) (S) (x6 . . We now use the result of ﬁberwise duality transformation: (A on KA × D/(hA · g)) ≡ (B on KB × D/(hB · g)) by choosing A on KA to be M-theory on S 1 . Sen: Duality Symmetries 317 propagating in the bulk [136]! It has been suggested that this physical separation could be as large as a millimeter [99]! This limit comes from the analysis of the ﬁfth force experiment. 10 Let us express this as (I1 · σ) · I4 where I1 changes the sign of x . and g to be I4 . hB to be (−1)FL (this can be shown to be the image of hA in the type IIA string theory). Many of the listed duality conjectures involving M-theory (in fact all except the ﬁrst one) can be derived by ﬁberwise duality transformation [98]27 . since gauge ﬁelds live on the boundary of S 1 /Z2 and hence do not get aﬀected by the existence of this extra dimension. D to be T 4 spanned by x6 . 27 The duality between E × E heterotic string theory and M-theory on S 1 can be 8 8 “derived” from other known duality conjectures by taking the inﬁnite radius limit of a lower dimensional duality relation [75].A. and I4 changes the sign of (x6 . Thus we get the duality M-theory on (S 1 × T 4 /I1 · σ · I4 ) ↔ IIA on T 4 /(−1)FL · I4 . Let us for example consider the duality (M theory on T 5 /Z2 ) ↔ (type IIB on K3). Thus we get the duality: (M-theory on T 5 /Z2 ) ↔ (IIB on K3). . x10 ) on T 5 . x9 . . No such direct limit comes from the inverse square law of gauge interaction. and • (−1)FL · I4 into I4 . Thus the theory on the right is dual to type IIB on T 4 /I4 . . . we should have inverse cube law for the gravitational force instead of inverse square law.

Johnson and J. gr-qc/9707012. B [Proc. M. Strominger. J. Sen. M. Vafa [hep-th/9702201]. Lett. B 78 (1978) 97. E. J. Polchinski [hep-th/9411028]. Schwarz and E. Porrati and E. C. Nucl. Rabinovici. M. E. Phys. 259 (1995) 213 [hep-th/9412184]. W. B 438 (1995) 109 [hep-th/9410167]. J. Gauge Theory and Strings This duality was ﬁrst conjectured in [96. Lectures on String Theory (Springer. M. Douglas [hep-th/9610041]. B 357 (1995) 545 [hep-th/9506194]. S. Phys. Witten. Forste and J. Duﬀ and R. hep-th/9712004]. Nucl. Hull and P. 135]. Phys. Witten. B 443 (1995) 85 [hep-th/9503124]. Proc. M. B. Dabholkar. Khuri and J. Rep. Sen. J. Polchinski [hep-th/9602052]. Harvey and A. Hull. Phys. B 460 (1996) 525 [hep-th/9510169]. M. Witten. Vols. Louis [hep-th/9801132]. Mod. Duﬀ. Phys. [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] . As in the case of type IIA string theory. Lu. Phys. Sup. Nucl. Lett. 134. A 9 (1994) 3707 [hep-th/9402002]. M. Nucl. Lett. hep-th/9710246]. Polchinski [hep-th/9611050]. Kiritsis [hep-th/9708130]. Superstring Theory. A. Phys.318 Unity from Duality: Gravity. Polchinski and E. Int. Phys. Rep. E. B 55 (1) [hep-th/9607201]. 1986). Duﬀ [hep-th/9611203]. A. Phys. edited by D. Green. R. de Wit and J. S. Theisen. Phys. C. Nucl. Klemm [hep-th/9705131]. Nucl. Nucl. C. P. Duﬀ. Rev. Witten and D. B 449 (1995) 535 [hep-th/9504047]. Suppl. Giveon. J. we can get non-perturbative enhancement of gauge symmetries in M-theory when the compact manifold develops singularities [19. Townsend [hep-th/9612121. Phys. J. Schwarz. 68 (1996) 1245 [hep-th/9607050]. J. Lerche [hep-th/9611190. Khuri. 1 and 2 (Cambridge University Press. Polchinski. References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] A. Trigiante [hep-th/9801144]. Mod. Townsend.] 45B. Nucl. Phys. The extra massless states required for this symmetry enhancement come from membranes wrapped around the collapsed two cycles of the singular manifold. J. Chaudhuri. Mukhi [hep-ph/9710470]. Phys. M. 1989). Lust and S. B 442 (1995) 47 [hep-th/9501030]. Olive. 244 (1994) 77 [hep-th/9401139]. Phys.C (1996) 59 [hep-th/9512043] and references therein. P. C. Nucl. Fre. B 357 (1995) 307 [hep-th/9506160]. S. A. A. B 411 (1994) 473 [hep-th/9305142]. B 450 (1995) 103 [hep-th/9504027]. Phys. M-theory contains classical membrane and ﬁve-brane soliton solutions carrying electric and magnetic charges of (S) CMNP respectively [8]. Phys. 97]. Louis [hep-th/9612192].

Geometry and Dynamics of Magnetic Monopoles (Cambridge Univ. Gibbons and N. Osborn. J. Schwarz and P. Dabholkar. Phys. Font. Phys. de Roo. Ruiz. Lett. B 390 (1993) 3 [hep-th/9207016]. Lett. A 107 (1985) 21. Schwarz. Schroers. B 303 (1993) 22 [hep-th/9209016]. 1981). Phys. in “Superspace and Supergravity” (Cambridge Univ. P. G. A.F. Lykken. B 460 (1996) 335 [hep-th/9510135]. Nucl. A. Quevedo. Gibbons. Phys. Hockney and J. Math. D. Phil. Lett. S. B 329 (1994) 217 [hep-th/9402032]. B 360 (1995) 13 [hep-th/9508143]. B 375 (1992) 103 [hep-th/9109038]. Phys. J. Comm. Phys. Annals Phys. Nucl. Rev. Chaudhuri. Phys. B 226 (1983) 269. Gauntlett and J. A. Nucl. Phys. Comm. 75 (1995) 2264 [hep-th/9505054]. M. Lett. Olive. Lett. D 52 (1995) 7168 [hep-th/9506048]. S.A. Lust and F. Phys. Phys. Lett. Lett. Hitchin. M. Phys. 177 (1996) 775. B 404 (1993) 109 [hep-th/9207053]. J. 49 (1989) 713. S. Phys. West. B 83 (1979) 321. H. Nucl. B 375 (1996) 103 [hep-th/9601152]. Ferrara. Roy. Chaudhuri and J. Press). Sen. Sen. Nucl. A 315 (1985) 459. J. B 238 (1984) 181. Harvey and F. Phys. Julia. Ruback. Phys.J. 63 (1989) 478. Lett. Mod. A. Sen. Phys. G. Nucl. J. Phys. Schwarz. C. Lett. A. Witten. Rev. Press. Soc. Ibanez. Schwarz and A. E. Phys. A. Phys. C. Kounnas and M. Dabholkar and J. Nucl. Schwarz. Shapere. Hassan and A. Gauntlett. Schwarz. B 126 (1983) 301. Manton and B. Rev. Phys. B 411 (1994) 35 [hep-th/9304154]. Phys. J. D 43 (1991) 526. Schwarz [hep-th/9209125]. B 340 (1990) 33. Sen: Duality Symmetries [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] 319 [45] [46] [47] [48] [49] [50] [51] [52] [53] [54] [55] [56] [57] [58] [59] [60] [61] [62] [63] [64] [65] M. West. Nucl. Nucl. Phys. Polchinski. Phys. G. L. B 249 (1990) 35. B 411 (1994) 443 [hep-th/9305068]. Trans. . Blum. N. Ferrara and C. Phys. Lett. 225 (1993) 290. Howe and P. J. Phys. B 72 (1977) 117. Harvey. G. Phys. J. B 328 (1989) 406. 1980). Lett. A 6 (1991) 2677. Sov. Phys. Maharana and J. in “Uniﬁcation of Fundamental Particle Interactions” (Plenum. Nucl. B 312 (1993) 105 [hep-th/9305185]. Sen. 75 (1995) 4724 [hep-th/9510017]. S. Phys. B 255 (1985) 515. Das and S. Lett. B 400 (1993) 103 [hep-th/9205008]. A. Polchinski. Sen. Cremmer. B 333 (1994) 92 [hep-th/9401133]. Phys. E. Porrati. S. Phys. A 8 (1993) 2023 [hep-th/9303057]. Segal and A. Rey. Mathur. Phys. Montonen and D. Atiyah and N. Gibbons and P. J. Nucl. Lett. Math. G. Nucl. Phys. Lett. Phys. M. Harvey [hep-th/9407111]. Lett. Green and J. J. Manton. S. Trivedi and F. Phys. A. Lett. M. Wilczek. Rev. Nucl. Seiberg. S. S. B 122 (1983) 143. Lett. B. B 377 (1996) 67 [hep-th/9505187]. B 274 (1986) 183. J. Selby. Phys. B 303 (1988) 286. Lett. Terentev. Phys. N. Lond. Rev. Phys. Nucl. 115 (1988) 267. Phys. J. Mod. Phys. Sen. Lett. Nucl. Porrati. Kounnas. B 181 (1986) 263.

C. Ferrara. Giani and M. Phys. Lett. Harvey. Phys. Phys. Lett. Witten. Dasgupta and S. Gava. Kachru. P. Antoniadis. P. Phys. Phys. [95] P. Stern [hep-th/9705046]. P. 2 (1985) 293. B 255 (1985) 569. Phys. Nucl. A. Class. Leigh. A. Schimmtigk [hep-th/9609082]. Phys. Nucl. Lynker and R. [83] B. Phys. D 30 (1984) 325. Nucl. B 469 (1996) 415 [hep-th/9511088]. J. Sadov and C. Phys. B 465 (1996) 399 [hep-th/9512196]. B 474 (1996) 361 [hep-th/9604070]. Lett. [86] I. [82] G. B. Blum. [69] C. Aspinwall. Phys. [79] S. [92] A. Stern. Campbell and P. Nucl. Phys. Mod. B 455 (1995) 109 [hep-th/9507115]. Polchinski. Lerche and P. [88] N. Bershadsky. Phys. Vafa. B 477 (1996) 746 [hep-th/9604034]. Quant. Phys. B 461 (1996) 537 [hep-th/9510093]. [80] S. . Mukhi. Polchinski. Dai. Phys. [76] A. Nucl. Nucl. Schwarz. [84] B. Bershadsky. Vafa. Kaplunovsky. B 506 (1997) 223 [hep-th/9705030]. M. Nucl. Townsend. B 447 (1995) 62 [hep-th/9503057]. Phys. Quevedo. Mod. Nucl. [74] J. Lett. A 4 (1989) 2073. B 381 (1996) 427 [hep-th/9512138]. Sethi and M. Sen. Mayr. [78] A. Phys. Sen. Nucl. B 459 (1996) 537 [hep-th/9508155]. I. Aldazabal. [94] P. Suppl. Phys. Lerche. Vafa and E. Phys. Mayr. A 4 (1989) 2767. B 357 (1995) 71 [hep-th/9506110]. Mayr and C. Leigh and J. R. [73] M. B 460 (1996) 506 [hep-th/9510209]. Nucl. [96] K. Phys. Pernici. B 434 (1995) 179 [hep-th/9408083]. V. Klemm. Nucl. Namazie. B 460 (1996) 541 [hep-th/9511030]. Nucl. de Wit. Nucl. B 426 (1994) 19 [hep-th/9407087]. B 367 (1996) 97 [hep-th/9510086]. Rev. Rev. Nucl. B 450 ((1995) 69 [hep-th/9605105]. Proc. Vafa. Rev. [77] C. Nucl. Witten [hep-th/9507050]. [68] A. Phys. Lerche. B 361 (1995) 59 [hep-th/9505162]. [81] A. Phys. B 357 (1995) 313 [hep-th/9506112]. Seiberg and E. Theisen. Phys. [87] S. Phys. Lett. L. [71] M. E. Nucl. Lett. R. B 398 (1997) 47 [hep-th/9607145]. W. West. [90] F. Klemm. Narain and T. Sadov and C. Phys. Witten. Phys. Kachru and C. Grav. Vafa. W. Taylor. J. [89] P. B 431 (1994) 484 [hep-th/9408099]. Warner. B 350 (1995) 184 [hep-th/9501068]. Sen [hep-th/9609176]. M. B 463 (1996) 420 [hep-th/9511222]. Strominger and C. [93] J. Lett. [72] E.320 Unity from Duality: Gravity. Ibanez and F. 46 (1996) 30 [hep-th/9508154]. Hunt. A. van Proeyen. Lett. Nucl. Nucl. Vafa and N. B 243 (1984) 112. Phys. Phys. Nucl. D 54 (1996) 2964 [hep-th/9510229]. Phys. Phys. Klemm. Nucl. Phys. Nucl. Rev. B 387 (1996) 492 [hep-th/9607082]. Hunt and R. W. Vafa. B 463 (1996) 398 [hep-th/9510225]. Sen. [75] A. Phys. Sethi and M. [91] S. Porrati. Lauwers and A. Phys. [67] M. Horava and E. Phys. [85] V. Nucl. J. Huq and M. Sen. Lett. Phys. D 50 (1994) 6041 [hep-th/9407031]. Vafa. K. B 475 (1996) 94 [hep-th/9603142]. Font. D 53 (1996) 2874 [hep-th/9711026]. Schimmrigk. Witten. Lett. Phys. Phys. B 463 (1996) 435 [hep-th/9512078]. Louis and S. Gauge Theory and Strings [66] J. V. [70] S. A.

J. Math. Phys. Lett. Phys. S. Phys. 80 (1981) 575. J. van Niewenhuizen. [112] E. Gross. Phys. Phys. 1987). Julia. B 463 (1996) 383 [hep-th/9512219]. B 110 (1982) 54. T. Zumino. Phys. Phys. M. Witten. Sen. Mandelberg. Morrison and A. [117] A. D 54 (1996) 1667 [hep-th/9601038]. Friedan and E. van Proeyen. Phys. B 279 (1987) 369. de Wit. Dixon. Nucl. B 493 (1997) 73 [hep-th/9606036]. Math. B. J. Phys. [113] M. Nucl. Douglas and J. Sarmadi and E. Phys. Rohm. B 299 (1988) 613. de Wit and L. Schwarz. Phys. D. E. Math. Phys. Witten [hep-th/9507121]. Nucl. [114] D. Lett. Harvey. Cremmer and B. 133 (1990) 163. [103] H. Horava. Lett. B 121 (1977) 393. D 24 (1981) 3065. B 151 (1985) 21. hep-th/9604198]. de Wit. P. Phys. Goddard. [116] C. Phys. Phys. Moore [hep-th/9603167]. Nicolai. Sen: Duality Symmetries 321 [97] E. C. Nucl. E. Phys. B 158 (1979) 189. Corrigan and P. [118] M. Comm. Phys. A. B. de Roo. B 231 (1989) 251. Nucl. B 255 (1985) 93. G. Diaconescu. Rev. Comm. 91 (1983) 235. [120] M. [109] B. Sagnotti. Cremmer. Polchinski. S. Callias. Nucl. Nucl. Lett.. K. Phys. Math. Rev. Phys. Greene [hep-th/9702155] and references therein. Comm. Martinec and R. Phys. [107] K. Lett. A. V. Ferrara and A. [119] E. B 451 (1995) 109 [hep-th/9504145]. Nucl. Nucl. Gaillard and B. B 74 (1978) 61. Ferrara. B 80 (1978) 48. B 193 (1981) 221. Greene. G. B 376 (1992) 365. Phys. Johnson and R. B 245 (1984) 89. Polchinski. B. B 250 (1985) 385. Nucl. Weinberg. E. Phys. M. Phys. H. Lett. 54 (1985) 502. Phys. Nucl. Manton. Phys. Girardello. K. Scherk and S. Bianchi. [99] E. Phys. Comm. de Wit and A. Cremmer. Nucl. Phys. B 245 (1984) 89. Ferrara. in “Strings 89”. Taubes. van Proeyen. B 451 (1995) 525 [hep-th/9505073]. Strominger. Phys.A. Narain. World Scientiﬁc (1989). J. Pradisi and A. Derendinger. B 327 (1989) 461. L. Bagger and E. Martinec. Nucl. Strominger. Math. Phys. Rev. [100] A. Lett. Phys. Pradisi and A. Phys. Ferrara. Witten. [98] A. Nucl. Strominger. B 109 (1982) 444. D. [111] J. Sagnotti. Phys. D 55 (1997) 6382 [hep-th/9610140]. Scherk and B. Lett. de Wit and P. Phys. Lett. Hull and P. Townsend. Nucl. . Phys. E. Phys. Galicki. Kounnas. B 451 (1995) 96 [hep-th/9504090]. B 169 (1986) 41. B 195 (1982) 97. van Proeyen. Nicolai. Phys. Phys. de Wit and H. Phys. [110] B. Phys. Bergshoeﬀ. B 217 (1983) 143. [102] A. C. B 120 (1983) 105. M. B 159 (1979) 141. [106] C. Lett. [101] E. M. Julia. Mod. B 267 (1986) 75. Rev. Dixon. G. B. Nucl. [108] T. Banks and L. Chamseddine. Phys. Open Strings and their Symmetry Groups (Talk at Cargese Summer Inst. Phys. D. Gomis [hep-th/9712230]. S. Nucl. D 20 (1979) 936. Nucl. B 55 (1997) 6423 [hep-th/9606165]. Phys. M. B 222 (1983) 1. A 11 (1996) 1339 [hep-th/9603113]. Zumino. Myers. [104] E. J. Nucl. E. B. Douglas [hep-th/9612126] published in JHEP electronics journal. de Wit and A. Manton. de Roo and B. Sagnotti. Nucl. Lett. Nucl. Douglas [hep-th/9512077. Caceres. Narain. B 216 (1989) 59. C. E. Phys. Lett. B 276 (1992) 333. B 307 (1988) 93. Lett. Witten. Nucl. Nucl. Rev. B 208 (1982) 323. Nucl. Gimon and J. Douglas and G. Strominger. Bergshoeﬀ. [115] M. 62 (1978) 213. 108 (1987) 117. Kaplunovsky and M. Rev. Phys. Chapline and N. Banks. B 256 (1985) 253. B 149 (1984) 117. Green and J. B. [105] N. Cremmer and B. Phys. Phys. Comm. E. Phys.

B 320 (1989) 135. Phys. Cardy. Sen [hep-th/9707123] published in JHEP electronic journal. Nucl. Phys. Phys. Gopakumar and S. . R. Alvarez-Gaume. Scherk and D. Pope. [123] P. B 65 (1976) 282. Phys. Phys. Horowitz. Phys. Johnson. Sonneschein. Lett. Nucl. B. Nilsson and C. Mukhi. J. Grav. S. S. J. Duﬀ. Duﬀ. B. Dasgupta and S. Park [hep-th/9611119]. Ann. B 305 (1988) 545. [130] M. Phys. J. Gaberdiel. Phys. Gliozzi. Duﬀ. Witten. Lett. B 387 (1996) 71 [hep-th/9607019]. Schwarz. S. B 258 (1985) 46. Rev. Class. (NY) 185 (1988) 330. Townsend.322 Unity from Duality: Gravity. S. M. Phys. Witten. Phys. C. J. B 479 (1996) 260 [hep-th/9607057]. [137] A. Hauer and B. Rev. [122] E. P. Nucl. B 489 (1997) 139 [hep-th/9611186]. Nicolai. Minasian [hep-th/9602102]. B 481 (1996) 215 [hep-th/9605200]. Lett. Duﬀ. Nucl. [131] K. D 55 (1997) 7345 [hep-th/9702165]. Phys. Aharony. R. Phys. B 31 (1971) 86. Johansen. B 452 (1995) 261 [hep-th/9506126]. Nucl. Phys. L. Dabholkar and J. Nucl. Phys. Ramond. Phys. B 493 (1997) 177 [hep-th/9611222]. Ferrara and T. K. Phys. Phys. Hoppe and H. Nucl. Lett. Phys. Phys. A. D 4 (1971) 1109. Leigh. [133] M. Yankielowicz and S. Nucl. Sen. Morrison. Zaﬀaroni. Nucl. [125] F. Zwiebach [hep-th/9709013]. Intrilligator. O. B 498 (1997) 135 [hep-th/9702061]. D. Nucl. P. B 460 (1996) 489. Seiberg and E. I. de Wit. (NY) 134 (1981) 85. Witten. Aspinwall [hep-th/9611137] and references therein. [132] A. Schwarz. [135] A. D. B 270 (1986) 186. R. Khuri and R. 130 (1986) 1. B 447 (1995) 261 [hep-th/9505053]. Nucl. Nucl. Polchinski. Phys. Taylor. Phys. [139] J. Math. Phys. Nicolai. Park. B 463 (1996) 435 [hep-th/9601036]. Cadavid. Phys. Blum and A. Berkooz. Candelas. 5 (1988) 189. Nucl. M. Witten. Nucl. B 394 (1997) 302 [hep-th/9607041]. Freedman. Phys. N. Mukhi. Phys. Sezgin and P. Vafa and E. T. Sadov and C. R. B 387 (1996) 735 [hep-th/9605131]. Morrison [hep-th/9404151]. A. Phys. Nucl. B 471 (1996) 195 [hep-th/9603150]. J. Gross. Antoniadis. Ferrara. B 189 (1987) 75. Alvarez-Gaume and D. B 122 (1977) 253. B 471 (1996) 135 [hep-th/9602070]. Phys. Gaberdiel and B. [124] A. Nucl. J. Phys. 80 (1981) 443. Bershadsky. Rev. B. Gauge Theory and Strings [121] L. Nucl. M. Duﬀ and R. A. B 436 (1995) 507 [hep-th/9406198]. Minasian and E. [138] M. [136] E. D’Auria and S. de Wit. Aspinwall and M. Phys. Witten. [134] E. E. J. Comm. [129] E. Ann. B 385 (1996) 125 [hep-th/9606044]. Phys. B 475 (1996) 115 [hep-th/9605184]. Liu and R. Vafa. Bergshoeﬀ. Aspinwall and D. Lett. Lett. Phys. Gimon and C. Ceresole. Minasian. M. D 3 (1971) 2415. Nucl. M. Theisen. Phys. Freedman and S. B 500 (1997) 3 [hep-th/9703166]. [126] P. E. Phys. Phys. M. Zwiebach [hep-th/9801205]. R. Nucl. Olive. Ferrara [hep-th/9506144]. Phys. Luscher and H. Neveu and J. G. Kachru. Mukhi. A. Minasian. Lett. Witten. [128] P. B 479 (1996) 285 [hep-th/9606176]. [127] M. B 395 (1997) 36 [hep-th/9608186]. J. Nucl. Strominger and E. Nucl. Quant. Rep. Witten. V.

MALDACENA Institute for Advanced Study. U.S.A. NJ. . Princeton.LECTURE 4 LES HOUCHES LECTURES ON LARGE N FIELD THEORIES AND GRAVITY J.

. . . . . . . .Contents 1 General introduction 325 2 The correspondence 330 2. . . . . . . . . . . . .2 Holography . . . . . . . . . .1 The ﬁeld ↔ operator correspondence . . . .2 Other branes ending on the boundary . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 Two-point functions . . . .1 Wilson loops and minimum surfaces . . . . . . . 369 6. . . . . . . . . . . . . . . 368 6 Theories at ﬁnite temperature 369 6. . . . . . . . . . . . . . . . 355 4. . . . . . . . . . . . . . . . 342 3. . . 372 Ë . 363 5. . . 336 2. . . . 351 4 Correlation functions 353 4.2 Thermal phase transition . . . . . . . . . . . . . . . . . .2 Three-point functions . . . . . . . . 359 5 Wilson loops 362 5. . 338 /CFT correspondence 341 3 Tests of the 3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 Construction . . . . . . . . . . . . . . . . . .1 The spectrum of chiral primary operators . . . . . . .2 Matching of correlation functions and anomalies . . .

This is explained simply by assuming that the mass and angular momentum come from a rotating.LES HOUCHES LECTURES ON LARGE N FIELD THEORIES AND GRAVITY J. Springer-Verlag 2002 . c EDP Sciences. Maldacena Abstract We describe the holographic correspondence between ﬁeld theories and string/M theory. We describe the main results that have been derived from the correspondence in the regime that the ﬁeld theory is approximated by classical or semiclassical gravity. the mass of the lightest hadron with a given spin obeys a relation like m2 ∼ T J 2 + const. The string idea described well some features of the hadron spectrum. These lecture notes are based on the Review written by Aharony et al. relativistic string of tension T . [1]. String theory was discovered in an attempt to describe the large number of mesons and hadrons that were experimentally discovered in the 1960’s. For example. Even though though string theory is normally used as a theory of quantum gravity. The idea was to view all these particles as diﬀerent oscillation modes of a string. 1 General introduction These lecture notes are taken out of the review [1]. We review the background for this correspondence and discuss its motivations and the evidence for its correctness. focusing on the relation between compactiﬁcations of string/M theory on Anti-de Sitter spaces and conformal ﬁeld theories. A more complete set of references is given there. We focus on the case of the N = 4 supersymmetric gauge theory in four dimensions. It was later discovered that hadrons and mesons are actually made of quarks and that they are described by QCD. it is not how string theory was originally discovered. This work is supported by the Packard foundation and the DOE grant DE-FGO291ER40654.

Gauge Theory and Strings QCD is a gauge theory based on the group SU (3). like any string theory. A simple example. It is quite hard to ﬁnd quantum ﬁeld theories that are conformally invariant. It was suggested by ’t Hooft that the theory might simplify when the number of colors N is large [7]. Strings are not consistent in four ﬂat dimensions. as explained in the next section. If the case with N = 3 is similar to the case with N = ∞ then this explains why the string model gave the correct relation between the mass and the angular momentum. the diagrammatic expansion of the ﬁeld theory suggests that the large N theory is a free string theory and that the string coupling constant is 1/N . If this is the case. is very general. reviewed below. The hope was that one could solve exactly the theory with N = ∞. why do we say that the string moves in ﬁve dimensions? The reason is that. One possible approach is to use numerical simulations on the lattice. including quantum gravity. The ’t Hooft argument. string theory in some backgrounds. This is sometimes stated by saying that quarks have three colors. which will be the main example . this theory will contain gravity. is equivalent (dual) to a ﬁeld theory. In particular we could consider a gauge theory where the coupling does not run (as a function of the energy scale). as described in detail below. Indeed. At low energies QCD becomes strongly coupled and it is not easy to perform calculations. and the gravitational theory will live in as many dimensions as the number of ﬁelds we have on the string. and then one could do an expansion in 1/N = 1/3. One might qualitatively think of this new ﬁeld as the “thickness” of the string. This is at present the best available tool to do calculations in QCD at low energies. In supersymmetric theories it is sometimes possible to prove exact conformal invariance. It is crucial then that the ﬁve dimensional geometry is curved. so that it can correspond to a four dimensional ﬁeld theory. The argument that gauge theories are related to string theories in the large N limit is very general and is valid for basically any gauge theory. QCD is asymptotically free. In this review we will study this correspondence between string theories and the large N limit of ﬁeld theories. sometimes called the “Liouville” ﬁeld [8]. if one wants to quantize a four dimensional string theory an anomaly appears that forces the introduction of an extra ﬁeld. Namely. Furthermore. so that the strings eﬀectively move in ﬁve dimensions. the theory is conformally invariant. We will see that the strings arising in the large N limit of ﬁeld theories are the same as the strings describing quantum gravity. Then.326 Unity from Duality: Gravity. In this way the large N limit connects gauge theories with string theories. This ﬁeld on the string worldsheet may be interpreted as an extra dimension. meaning that the eﬀective coupling constant decreases as the energy increases. so it suggests that diﬀerent kinds of gauge theories will correspond to diﬀerent string theories.

So. Now that we have added one more dimension it is not surprising any more to add ﬁve more to get to a ten dimensional space. which describe the oscillations of the branes. If they have zero spatial dimensions they are like ordinary localized. The Lagrangian of such theories is completely determined by supersymmetry. We said that superstrings move in ten dimensions. Here we have presented a very heuristic argument for this equivalence. so they carry two indices that run from one to N . Besides the gauge ﬁelds (gluons) this theory contains also four fermions and six scalar ﬁelds in the adjoint representation of the gauge group. particle-type soliton solutions. S 5 . the tension of all D-branes is proportional to 1/gs . They come in various dimensionalities. There is a global SU (4) R-symmetry that rotates the six scalar ﬁelds and the four fermions. If we have N coincident branes the open strings can start and end on diﬀerent branes. These symmetries of the ﬁeld theory should be reﬂected in the dual string theory. This in turn implies that the low energy dynamics is described by a U (N ) gauge theory. we conclude that N = 4 U (N ) Yang-Mills theory could be the same as ten dimensional superstring theory on AdS5 × S 5 [9]. where gs is the string coupling constant.J. 2) isometries: ﬁve dimensional Anti-de-Sitter space. D-branes are deﬁned in string perturbation theory in a very simple way: they are surfaces where open strings can end. These are called D-zero-branes. D-branes are solitons in string theory [10]. 2). analogous to the ’t Hooft-Polyakov [11. Anti-de Sitter space is the maximally symmetric solution of Einstein’s equations with a negative cosmological constant. Since the gauge theory has an SU (4) SO(6) global symmetry it is rather natural that the extra ﬁve dimensional space should be a ﬁve sphere. In this supersymmetric case we expect the strings to also be supersymmetric. If they have one extended dimension they are called Done-branes or D-strings. The relationship we described between gauge theories and string theory on Anti-de-Sitter spaces was motivated by studies of D-branes and black holes in strings theory. Four is the maximal possible number of supercharges for a ﬁeld theory in four dimensions. They are much heavier than ordinary fundamental strings when the string coupling is small. These open strings have some massless modes. including the usual Poincar´ transformations as well as scale transformae tions and special conformal transformations (which include the inversion symmetry xµ → xµ /x2 ). The conformal group in four dimensions is SO(4. is the supersymmetric SU (N ) (or U (N )) gauge theory in four dimensions with four spinor supercharges (N = 4). Maldacena: Lectures on Large N Field Theories and Gravity 327 in this review. and their fermionic partners. The simplest way for this to happen is if the ﬁve dimensional geometry has these symmetries. In fact. 12] monopole in gauge theories. Locally there is only one space with SO(4. later we will be more precise and give more evidence for this correspondence. D-p-branes are charged under . a gauge ﬁeld living on the brane. or AdS5 .

More generally. On the other hand low energy excitations on the branes are governed by the Yang-Mills theory. These solutions are very similar to extremal charged black hole solutions in general relativity. and then agreement was shown for near extremal black holes [18. Gauge Theory and Strings p+1-form gauge potentials. and they are part of the massless closed string modes. Indeed it is possible to ﬁnd solutions of the supergravity equations carrying these ﬂuxes. in turn. which turns out to imply that gs N should be very large. it becomes natural to conjecture that Yang-Mills theory at strong coupling is describing the near horizon region of the black brane. If we consider a set of N coincident D-3-branes the near horizon geometry turns out to be AdS5 × S 5 . As an object is brought closer and closer to the black brane horizon its energy measured by an outside observer is redshifted. Like black holes they contain event horizons. in the same way that a 0-brane (particle) can be charged under a one-form gauge potential (as in electromagnetism). the low energy dynamics on their worldvolume is governed by a U (N ) gauge theory with N = 4 supersymmetry [13]. we will see that correlation functions in the gauge theory can be calculated using the string theory (or gravity for large gs N ) description. So. and the energy seems to be very small. On the other hand. due to the large gravitational potential. except that in this case they are black branes with p extended spatial dimensions. These calculations. while the low-energy gravitational description is perturbatively valid when the radius of curvature is much larger than the string scale. These two pictures of D-branes are perturbatively valid for diﬀerent regimes in the space of possible coupling constants. First black hole entropy for extremal black holes was calculated in terms of the ﬁeld theory in [17]. These p + 1-form gauge potentials have p + 2-form ﬁeld strengths. were inspired by similar calculations for coincident D1-D5 branes. and this ﬂux in turn contributes to the stress energy tensor so the geometry becomes curved. by considering the propagation of . If we now add D-branes they generate a ﬂux of the corresponding ﬁeld strength. It was noticed there that the calculation done using gravity and the calculation done using super Yang-Mills theory agreed. Supergravity is the low-energy limit of string theory. In this case the near horizon geometry involves AdS3 × S 3 and the low energy ﬁeld theory living on the D-branes is a 1+1 dimensional conformal ﬁeld theory. which belong to the supergravity (SUGRA) multiplet containing the massless ﬁelds in ﬂat space string theory (before we put in any D-branes). Perturbative ﬁeld theory is valid when gs N is small. and it is believed that these solutions may be extended to solutions of the full string theory. In this D1-D5 case there were numerous calculations that agreed between the ﬁeld theory and gravity. whose geometry is AdS5 × S 5 . The ﬁrst indications that this is the case came from calculations of low energy graviton absorption cross sections [14–16].328 Unity from Duality: Gravity.19] and for absorption cross sections [20–22].

One of the main points of these lectures will be that the strings coming from gauge theories are very much like the ordinary superstrings that have been studied during the last 20 years. by taking the limit N → ∞. then the U (N ) Yang-Mills theory gives a non-perturbative deﬁnition of string theory on AdS. However. since any string theory includes gravity. 26] for reviews. And. In some sense the ﬁeld theory (or at least the set of local observables in the ﬁeld theory) lives on the boundary of spacetime. but a light ray can go to the boundary and come back in ﬁnite time. see [25. This is a topological theory in three dimensions that induces a normal (non-topological) ﬁeld theory on the boundary. The radius of curvature of Anti-de Sitter space depends on N so that large N corresponds to a large radius of curvature. The asymptotically AdS case as well as the asymptotically ﬂat cases are special in the sense that one can choose a natural time and an associated Hamiltonian to deﬁne the quantum theory. See also [2. but the spacetime could have any topology as long as it is asymptotically ﬂat. but it is certainly well-deﬁned at inﬁnity. So in the end we claim that there is an equivalence between a gravitational theory and a ﬁeld theory. Maldacena: Lectures on Large N Field Theories and Gravity 329 particles between diﬀerent points in the boundary of AdS. by taking N to be large we can make the curvature as small as we want. 24]. So this is really a sum over all spacetimes that are asymptotic to AdS at the boundary. Notice that when we say that the theory includes “gravity on AdS” we are considering any ﬁnite energy excitation. The boundary is at an inﬁnite spatial distance. Since black holes might be present this time coordinate is not necessarily globally well-deﬁned. The theory in AdS includes gravity. If we assume that the conjecture we made above is valid. 3] for earlier hints of the correspondence. Thus. In some sense the situation is similar to the correspondence between three dimensional Chern-Simons theory and a WZW model on the boundary [27]. even black holes in AdS. A theory which includes gravity is in some sense topological since one is integrating over all metrics and therefore the theory does not depend on the metric. One could argue that in general any quantum gravity theory in AdS deﬁnes a conformal ﬁeld theory (CFT) “on the boundary”. where asymptotic ﬂatness is required. in a quantum gravity theory we do not have any local observables. The only particular feature is that they are moving on a curved geometry (anti-de Sitter space) which has a boundary at spatial inﬁnity. Massive particles can never get to the boundary. Similarly. Supergravities on AdS spaces were studied very extensively. the mapping between the gravitational and ﬁeld theory degrees of freedom is quite non-trivial since the ﬁeld theory lives in a lower dimension.J. the points where operators are inserted [23. This is analogous to the usual ﬂat space discussion of quantum gravity. we can extract the (ten .

and their low-energy eﬀective Lagrangian is that of type IIB supergravity. The closed string massless states give a gravity supermultiplet in ten dimensions. closed strings and open strings. The closed strings are the excitations of empty space and the open strings end on the D-branes and describe excitations of the D-branes. The D3 branes are extended along a (3 + 1) dimensional plane in (9 + 1) dimensional spacetime.330 Unity from Duality: Gravity. Let us start with type IIB string theory in ﬂat. It was suggested [29. plus some higher derivative corrections.1) Sbulk is the action of ten dimensional supergravity.1) involves only the massless ﬁelds but it takes into account the eﬀects of integrating out the massive ﬁelds. This “holographic” principle comes from thinking about the Bekenstein bound which states that the maximum amount of entropy in some region is given by the area of the region in Planck units [31]. The fact that the ﬁeld theory lives in a lower dimensional space blends in perfectly with some previous speculations about quantum gravity. (2. ten dimensional Minkowski space. then only the massless string states can be excited. a procedure which is in principle (but not in detail) similar to the one used in matrix theory [28]. The reason for this bound is that otherwise black hole formation could violate the second law of thermodynamics. It is not renormalizable (even for the ﬁelds on the brane). 2 The correspondence In this section we will present an argument connecting type IIB string theory compactiﬁed on AdS5 ×S 5 to N = 4 super-Yang-Mills theory [9]. Consider N parallel D3 branes that are sitting together or very close to each other (the precise meaning of “very close” will be deﬁned below). in the sense that physics in some region can be described by a theory at the boundary with no more than one degree of freedom per Planck area. The open string massless states give an N = 4 vector supermultiplet in (3 + 1) dimensions.36]. Note that the Lagrangian (2. We will see that the correspondence between ﬁeld theories and string theory on AdS space (including gravity) is a concrete realization of this holographic principle. String theory on this background contains two kinds of perturbative excitations. Gauge Theory and Strings dimensional string theory) ﬂat space physics. and it . The complete eﬀective action of the massless modes will have the form S = Sbulk + Sbrane + Sint . Other reviews of this subject are [1. 30] that quantum gravity theories should be holographic. energies lower than the string scale 1/ls . If we consider the system at low energies. and we can write an eﬀective Lagrangian describing their interactions. 32–35]. and their low-energy eﬀective Lagrangian is that of N = 4 U (N ) super-Yang-Mills theory [13.

Dbranes are massive charged objects which act as a source for the various supergravity ﬁelds. and it contains the N = 4 super-Yang-Mills Lagrangian plus some higher derivative corrections. Next.J. can be expanded in a similar way. the supergravity theory in the bulk becomes free. In this limit the coupling κ ∼ gs α 2 → 0. leaving just the pure N = 4 U (N ) gauge theory in 3 + 1 dimensions.2) where we have written the metric as g = η + κh. involving other ﬁelds. We indicate explicitly the dependence on the graviton. In order to see more clearly what happens in this low energy limit it is convenient to keep the energy ﬁxed and send ls → 0 (α → 0) keeping all the dimensionless parameters ﬁxed. all interaction terms proportional to κ drop out. This is the well known fact that gravity becomes free at long distances (low energies). introducing the background metric for the brane [37]. for example terms of the form α 2 Tr(F 4 ). Similarly.e. Schematically we have Sbulk ∼ 1 2κ2 √ gR ∼ (∂h)2 + κ(∂h)2 h + · · · . (2. in this low energy limit we have two decoupled systems. the interaction Lagrangian Sint is proportional to positive powers of κ. we integrate out all massive degrees of freedom but we do not integrate out the massless ones. Maldacena: Lectures on Large N Field Theories and Gravity 331 should only be understood as an eﬀective description in the Wilsonian sense. In addition all the higher derivative terms in the brane action vanish. so that the interaction Lagrangian relating the bulk and the brane vanishes. And. plus some interactions which are proportional to positive powers of the square root of the Newton constant. of . We can expand the bulk action as a free quadratic part describing the propagation of free massless modes (including the graviton). including the string coupling constant and N . On the one hand we have free gravity in the bulk and on the other hand we have the four dimensional gauge theory. The brane action Sbrane is deﬁned on the (3 + 1) dimensional brane worldvolume. Finally. i. So. Sint describes the interactions between the brane modes and the bulk modes. but the other terms in the Lagrangian. We can ﬁnd a D3 brane solution [38] of supergravity. The leading terms in this interaction Lagrangian can be obtained by covariantizing the brane action. If we take the low energy limit. which is known to be a conformal ﬁeld theory. we consider the same system from a diﬀerent point of view.

In conclusion. In the near horizon region. one is free bulk supergravity and the second is the near horizon region of the geometry. r R. We see that both from the point of view of a ﬁeld theory of open strings living on the brane. the excitations that live very close to r = 0 ﬁnd it harder and harder to climb the gravitational potential and escape to the asymptotic region.332 the form Unity from Duality: Gravity. we are led to the conjecture that N = 4 U (N ) super-Yang-Mills theory in . the energy Ep of an object as measured by an observer at a constant position r and the energy E measured by an observer at inﬁnity are related by the redshift factor E = f −1/4 Ep . 1 2 3 5 R2 r (2. (2. where ω is the energy. 1 2 3 5 F5 = (1 + ∗)dtdx1 dx2 dx3 df −1 . Gauge Theory and Strings ds2 = f −1/2 (−dt2 + dx2 + dx2 + dx2 ) + f 1/2 (dr2 + r2 dΩ2 ) . In the low energy limit these two types of excitations decouple from each other. Note that since gtt is non-constant.4) This means that the same object brought closer and closer to r = 0 would appear to have lower and lower energy for the observer at inﬁnity. This can be understood from the fact that in this limit the wavelength of the particle becomes much bigger than the typical gravitational size of the brane (which is of order R). we have two decoupled theories in the low-energy limit. Now we take the low energy limit in the background described by equation (2. f =1+ R . or we can have any kind of excitation that we bring closer and closer to r = 0. We can have massless particles propagating in the bulk region with wavelengths that becomes very large. The bulk massless particles decouple from the near horizon region (around r = 0) because the low energy absorption cross section goes like σ ∼ ω 3 R8 [14. There are two kinds of low energy excitations (from the point of view of an observer at inﬁnity). Similarly. it is natural to identify the second system which appears in both descriptions.3). So.3) R4 ≡ 4πgs α 2 N . 15]. and from the point of view of the supergravity description. we can approximate f ∼ R4 /r4 . and the geometry becomes ds2 = dr2 r2 (−dt2 + dx2 + dx2 + dx2 ) + R2 2 + R2 dΩ2 . the low energy theory consists of two decoupled pieces. r4 4 (2. In both cases one of the decoupled systems is supergravity in ﬂat space.5) which is the geometry of AdS5 × S 5 . Thus.

The U (1) vector supermultiplet includes six scalars which are related to the center of mass motion of all the branes [39]. Depending on this choice we could have a correspondence to an SU (N ) or a U (N ) theory.6) This can also be seen by considering a D3 brane sitting at r. From the AdS point of view these zero modes live at the boundary. so there are no decoupled modes. as we did when we discussed the ﬁeld theory living on the brane. the bulk AdS theory is describing the SU (N ) part of the gauge theory. which is the way energies are measured in the ﬁeld theory. so this quantity should remain ﬁxed in the decoupling limit. It is then convenient to deﬁne a new variable U ≡ r/α . The U (1) center of mass degree of freedom is related to the topological theory of B-ﬁelds on AdS [40]. is proportional to U = r/α . (−dt2 + dx2 + dx2 + dx2 ) + 4πgs N ds2 = α √ 1 2 3 5 U2 4πgs N (2. Suppose that we take α → 0. We could be a bit more precise about the near horizon limit and how it is being taken. This mass. In the dual string theory all modes interact with gravity. we need to take r → 0 keeping r/α ﬁxed. so that the metric becomes dU 2 U2 + dΩ2 . . which correspond to the U (1) degrees of freedom mentioned above. the excitations in the asymptotic ﬂat space and the excitations in the near horizon region. Therefore. This implies that √ √ α Ep ∼ ﬁxed. 42]. so that we can consider arbitrary excited string states there. Since we want to keep ﬁxed the energy measured from inﬁnity.4) reduces to E ∼ Ep r/ α . We want to keep ﬁxed the energies of the objects in the throat (the near-horizon region) in string units. In fact we were not precise when we said that there were two sets of excitations at low energies. if one imposes local boundary conditions for these B-ﬁelds at the boundary of AdS one ﬁnds a U (1) gauge ﬁeld living at the boundary [41]. These modes living at the boundary are sometimes called singletons (or doubletons) [43–51]. There are also some zero modes which live in the region connecting the “throat” (the near horizon region) with the bulk. as is familiar in Chern-Simons theories [27.J. A U (N ) gauge theory is essentially equivalent to a free U (1) vector multiplet times an SU (N ) gauge theory. Maldacena: Lectures on Large N Field Theories and Gravity 333 3 + 1 dimensions is the same as (or dual to) type IIB superstring theory on AdS5 × S 5 [9]. When we take the α → 0 limit we want to keep the mass of the “W -boson” ﬁxed. This corresponds to giving a vacuum expectation value to one of the scalars in the Yang-Mills theory. up to some ZN identiﬁcations (which aﬀect only global issues). which is the mass of the string stretching between the branes sitting at r = 0 and the one at r. For small α (2. and it looks like we might or might not decide to include them in the AdS theory.

since the superconformal algebra has twice as many fermionic generators as the corresponding Poincare superalgebra. 52] τ≡ 4πi i θ χ = . the whole supergroup is the same for the N = 4 ﬁeld theory and the AdS5 × S 5 geometry. so there is a term of the form d4 xTr(φ2 )R in the Lagrangian. Since the isometries of AdS are in one to one correspondence with the generators of the conformal group of the ﬁeld theory. This formulation of the conjecture is more useful since in the global coordinates there is no horizon. so both sides of the conjecture have the same spacetime symmetries. Thus. This doubling of supersymmetries is viewed in the ﬁeld theory as a consequence of superconformal invariance. We have written the couplings in this fashion because both the gauge theory and the string theory have an SL(2. This is due to the fact that the scalars φI in the ﬁeld theory are conformally coupled. In fact.7) From the physics of D-branes we know that the Yang-Mills coupling is related to the string coupling through [10. Z) selfduality symmetry under which τ → (aτ + b)/(cτ + d) (where a. 2) for the case at hand. S5 F5 = N. When we put the ﬁeld theory on S 3 the Coulomb branch is lifted and there is a unique ground state. c. Due to the positive curvature of S 3 this leads to a mass term for the scalars [24]. The parameter N appears on the string theory side as the ﬂux of the ﬁve-form Ramond-Ramond ﬁeld strength on the S 5 . but we could also think of the conformal ﬁeld theory as deﬁned on S 3 × R by redeﬁning the Hamiltonian. + + 2 gYM 2π gs 2π (2. In the above derivation the ﬁeld theory is naturally deﬁned on R3. Gauge Theory and Strings Anti-de-Sitter space has a large group of isometries. We also have some supersymmetries. lifting the moduli space. we can conclude that this new Hamiltonian 1 (P0 + K0 ) can 2 be associated on AdS to the generator of translations in global time. This is the same group as the conformal group in 3 + 1 dimensions. We will discuss in more detail the matching between the two sides of the correspondence in Section 3. (2. where R is the curvature of the four-dimensional space on which the theory is deﬁned.334 Unity from Duality: Gravity. We also have an SO(6) symmetry which rotates the S 5 .1 . The number of supersymmetries is twice that of the full solution (2. the fact that the low-energy ﬁeld theory on the brane is conformal is reﬂected in the fact that the near horizon geometry is Anti-de-Sitter space.8) where we have also included the relationship of the θ angle to the expectation value of the RR scalar χ. This can be identiﬁed with the SU (4)R R-symmetry group of the ﬁeld theory. b. which is SO(4. d are .3) containing the asymptotic region [39].

J. R4 2 ∼ gs N ∼ gYM N 4 ls 1. 2. and it should also be a symmetry in the present context since all the ﬁelds that are being turned on in the AdS5 × S 5 background (the metric and the ﬁve form ﬁeld strength) are invariant under this symmetry. With these conventions GN ∼ 1/N 2 and α ∼ 1/ gs N . to set the radius of curvature to one. this is not really a parameter in the theory if we do not compare it to other scales in the theory. In this fashion we avoid any obvious contradiction due to the fact that the two theories look very diﬀerent. namely α . will be independent of gs N and will depend only on N . This can s be achieved by writing the metric as ds2 = R2 d˜2 . (2. Now. α will disappear from any ﬁnal physical quantity we compute in this theory. Maldacena: Lectures on Large N Field Theories and Gravity 335 integers with ad − bc = 1). In fact. Z) duality symmetries of type IIB string theory and N = 4 SYM was noted in [54–56]. and rewriting everything √ in terms of g . The analysis of loop diagrams in the ﬁeld theory shows that we can trust the perturbative analysis in the Yang-Mills theory when 2 gYM N ∼ gs N ∼ R4 4 ls 1. since only relative scales are meaningful. The two theories are conjectured to be exactly the same. the classical gravity description becomes reliable when the radius of curvature R of AdS and of S 5 becomes large compared to the string length. without including stringy eﬀects. This makes the correspondence both hard to prove and useful. This is the reason that this correspondence is called a “duality”.10) We see that the gravity regime (2. but not α and the radius of curvature independently. ˜ This implies that any quantity calculated purely in terms of the gravity solution. Z) is a conjectured strong-weak coupling duality symmetry of type IIB string theory in ﬂat space [53]. α corrections to the gravity results give corrections √ which are proportional to powers of 1/ gs N . Thus. (2. The string theory seems to have a parameter that does not appear in the gauge theory. but when one side is weakly coupled the other is strongly coupled and vice versa. as we can solve a strongly coupled gauge theory via classical supergravity. The connection between the SL(2. especially when one is doing gravity calculations.9) 2 2 Note that we need gYM N to be small and not just gYM . If gs > 1 we can perform an SL(2.10) and the perturbative ﬁeld theory regime (2. Z) .10) we implicitly assumed that gs < 1. However. let us address the question of the validity of various approximations. which sets the string tension and all other scales in the string theory. SL(2.9) are perfectly incompatible. On the other hand. only the ratio of the radius of curvature to α is a parameter. Notice that in (2.9. In fact. It is sometimes convenient.

. to have large N in order to have a weakly coupled supergravity description.336 Unity from Duality: Gravity. 2. In its weakest form the gravity description would be valid for large gs N .10) but with gs → 1/gs . which is then related to the expectation value of the dilaton. It is not a proof because we did not treat the string theory non-perturbatively (not even non-perturbatively in α ). So. but the gs corrections may not). we cannot get into the gravity regime (2. so the natural objects to consider are operators. We can consider adding the term d4 xφ0 (x)O(x) to the Lagrangian (for simplicity we assume that such a term was not present in the original Lagrangian. More precisely. One might wonder why the above argument was not a proof rather than a conjecture. For example. in N = 4 superYang-Mills we have a deformation by a marginal operator which changes the value of the coupling constant. 2. but not suﬃcient. black holes. So.9. The Yang-Mills theory is supposed to eﬀectively sum over all spacetimes which are asymptotic to AdS5 × S 5 . but the full string theory on AdS might not agree with the ﬁeld theory. A not so weak form would say that the conjecture is valid even for ﬁnite gs N . We could also consider diﬀerent forms of the conjecture.) and gs very large.8) to changing the coupling constant in the string theory. In the interior we can have all kinds of processes. In this case asymptotic ﬂatness is replaced by the asymptotic AdS behavior. . This is completely analogous to the usual conditions of asymptotic ﬂatness. since in that case the D-string becomes light and renders the gravity approximation invalid.1 The ﬁeld ↔ operator correspondence A conformal ﬁeld theory does not have asymptotic states or an S-matrix. Gauge Theory and Strings duality transformation and get conditions similar to (2. which is the most interesting one and which we will assume here. etc. According to the . but only in the N → ∞ limit (so that the α corrections would agree with the ﬁeld theory. otherwise we consider φ0 (x) to be the total coeﬃcient of O(x) in the Lagrangian). We can have black holes and all kinds of topology changing processes. 2. as long as spacetime is asymptotically ﬂat. D-branes. The expectation value of the dilaton is set by the boundary condition for the dilaton at inﬁnity. Even the topology of spacetime can change in the interior. gravitons. Changing the coupling constant in the ﬁeld theory is related by (2. In this conjecture the spacetime is only required to be asymptotic to AdS5 ×S 5 as we approach the boundary. let us denote by O the corresponding operator. Another way to 4 see this is to note that the radius of curvature in Planck units is R4 /lp ∼ N .10) by taking N small (N = 1. changing the gauge theory coupling constant corresponds to changing the boundary value of the dilaton. So. is that the two theories are exactly the same for all values of gs and N . it is always necessary.. highly excited fundamental string states. The strong form of the conjecture.

J.12). . Notice that φ0 is a function of the four variables parametrizing the boundary of AdS5 . that the associated operator O has dimension ∆ (2.11) is valid in general. we see that φ0 has dimensions of [length]∆−d which implies. The wave equation in Euclidean space for a ﬁeld of mass m has two independent solutions. 4 (2. the boundary condition on the ﬁeld in the right hand side of (2. The right hand side is the full partition function of string theory with the boundary condition that the ﬁeld φ has the value φ0 on the boundary of AdS. ) = d−∆ φ0 (x). as argued in [23. z)|z=0 = φ0 (x). which behave like z d−∆ and z ∆ for small z (close to the boundary of AdS).e. Maldacena: Lectures on Large N Field Theories and Gravity 337 discussion above. 24]. A more detailed derivation of this relation will be given in Section 4. (2. φ0 is an arbitrary function and we can calculate correlation functions of O by taking functional derivatives with respect to φ0 and then setting φ0 = 0. Therefore. Let us describe this more generally in AdSd+1 .11). including fermions and tensors on AdS space. in the coordinate system 2 ds2 = RAdS −dt2 + dx2 + · · · + dx3 + dz 2 1 3 · z2 More precisely. it is natural to propose that Ê e d4 xφ0 (x)O(x) CFT = Zstring φ(x. (2. A similar relation between ﬁelds on AdS and operators in the ﬁeld theory exists also for non-scalar ﬁelds. in order to get consistent behavior for a massive ﬁeld. Since φ is dimensionless.12) Therefore. z) z=0 = φ0 (x) .24].11) where the left hand side is the generating function of correlation functions in the ﬁeld theory.11) should in general be changed to φ(x. where ∆= d + 2 d2 + R 2 m2 .13) and eventually we would take the limit where → 0. for any ﬁeld φ. There is a relation between the mass of the ﬁeld φ and the scaling dimension of the operator in the conformal ﬁeld theory. through the left hand side of (2. it is natural to assume that this will change the boundary condition of the dilaton at the boundary of AdS to φ(x. i. A formula like (2. where we will verify that the two-point correlation function of the operator O behaves as that of an operator of dimension ∆ [23. each ﬁeld propagating on AdS space is in a one to one correspondence with an operator in the ﬁeld theory.

Since the correspondence described above matches the stress-energy tensor with the graviton. String theory provides such a theory. the only diagrams that contribute are the tree-level diagrams of the gravity theory (see for instance Fig. this implies that the AdS theory includes gravity.338 Unity from Duality: Gravity. and they should correspond to some conformal ﬁeld theory “on the boundary”. corresponding to a conformal ﬁeld theory in d spacetime dimensions (for d > 1). Fig. But if a new way of deﬁning quantum gravity theories comes along we could consider those gravity theories in AdS. this discussion can be extended to any AdSd+1 space. Feynman diagrams can be used to compute the interactions of particles in the bulk. In the limit where classical supergravity is applicable. we could consider backgrounds of string theory of the form AdS5 × M 5 where M 5 is any Einstein manifold [63–65]. Any local ﬁeld theory contains the stress tensor as an operator. 1). It should be a well deﬁned quantum theory of gravity since we should be able to compute loop diagrams. At tree level. 2. which sends a φ particle (a closed string state) into the bulk. . these diagrams and those related to them by crossing are the only ones that contribute to the four-point function.2 Holography In this section we will describe how the AdS/CFT correspondence gives a holographic description of physics in AdS spaces.11) by diﬀerentiating with respect to φ0 . Gauge Theory and Strings Correlation functions in the gauge theory can be computed from (2. Each diﬀerentiation brings down an insertion O. This method of deﬁning the correlation functions of a ﬁeld theory which is dual to a gravity theory in the bulk of AdS space is quite general. Similarly. 1. Correlation functions can be calculated (in the large gs N limit) in terms of supergravity Feynman diagrams. In particular. Here we see the leading contribution coming from a disconnected diagram plus connected pieces involving interactions of the supergravity ﬁelds in the bulk of AdS. and it applies in principle to any theory of gravity [24]. Depending on the choice of M 5 we get diﬀerent dual conformal ﬁeld theories.

30] (so that its entropy satisﬁes the Bekenstein bound). Maldacena: Lectures on Large N Field Theories and Gravity 339 Let us start by explaining the Bekenstein bound. which is smaller than the area of the boundary of the region (which by our assumption is smaller than the initial entropy). Suppose that we had a state with more entropy than Smax . the limit δ → 0 corresponds to going to the UV of the ﬁeld theory. When we are close to the boundary we could also use the Poincar´ e coordinates −dt2 + dx2 + dz 2 . the second law has been violated. So. For this purpose let us write the metric of AdS as ds2 = R2 − 1 + r2 1 − r2 2 dt2 + 4 (dr2 + r2 dΩ2 ) .14) In these coordinates the boundary of AdS is at r = 1. But if a black hole forms inside the region its entropy is just the area of its horizon. We saw above that when we calculate correlation functions we have to specify boundary conditions at r = 1 − δ and then take the limit of δ → 0. (2. has an inﬁnite number of degrees of freedom. it is hard to check what the number of degrees of freedom per Planck area is. then we show that we could violate the second law of thermodynamics. In the AdS/CFT correspondence we are describing physics in the bulk of AdS space by a ﬁeld theory of one less dimension (which can be thought of as living on the boundary). and the area of the boundary of AdS space is also inﬁnite. Thus. Note that this bound implies that the number of degrees of freedom inside some region grows as the area of the boundary of a region and not like the volume of the region. We can throw in some extra matter such that we form a black hole. However. (1 − r2 )2 (2.14) we see that its motion is independent of R in . So.15) ds2 = R2 z2 in which the boundary is at z = 0. If we consider a particle or wave propagating in (2. being conformal. which states that the maximum entropy in a region of space is Smax = Area/4GN [31]. so it looks like holography. It is clear by studying the action of the conformal group on Poincar´ coordinates that e the radial position plays the role of some energy scale.J. The entropy should not decrease. In standard quantum ﬁeld theories this is certainly not possible.15) or (2. since the theory. in order to compare things properly we should introduce a cutoﬀ on the number of degrees of freedom in the ﬁeld theory and see what it corresponds to in the gravity theory. since we approach the boundary when we do a conformal transformation that localizes objects in the CFT. which states that in a quantum gravity theory all physics within some volume can be described in terms of some theory on the boundary which has less than one degree of freedom per Planck area [29. where the area is that of the boundary of the region. Attempting to understand this behavior leads to the “holographic principle”.

This implies that a cutoﬀ at z∼δ (2.16) is called the UV-IR relation [66]. . since it does not seem to be saying much because in AdS space the volume and the boundary area of a given region scale in the same fashion as we increase the size of the region. We can estimate the number of degrees of freedom in the ﬁeld theory with a UV cutoﬀ δ. 4GN 4GN (2. Equation (2. if we are in Euclidean space and we have a wave that has some spatial extent λ in the x directions. One might get confused by the fact that the surface r = 1 − δ is really nine dimensional as opposed to four dimensional. we see that the theory on this nine dimensional surface becomes eﬀectively four dimensional. it will also have an extent λ in the z direction. the S 5 .14) the area in Planck units of the surface at r = 1 − δ.17) since the number of cells into which we divide the three-sphere is of order 1/δ 3 .340 Unity from Duality: Gravity. We get S ∼ N 2 δ −3 . remain constant. One could be a little suspicious of the statement that gravity in AdS is holographic. any ﬁeld theory in AdS would be holographic in the sense that the number of degrees of freedom within some (large enough) volume is proportional to the area (and also to the volume). with no factors of R (δ here is dimensionless. Furthermore. In fact. Consider the case of N = 4 SYM on a three-sphere of radius one. since these have a diﬀerent dependence on R. since we need to multiply the metric by a factor that goes to zero as we approach the boundary in order to deﬁne a ﬁnite metric for the four dimensional gauge theory. and then we can take AdS spaces of diﬀerent radii (corresponding to diﬀerent values of N in the SYM theory). is V 5 R3 δ −3 Area = S ∼ N 2 δ −3 . while the other ﬁve. and then we can ask whether the number of degrees of freedom goes like the volume or the area. in the ﬁeld theory it is measured in terms of the radius of the S 4 or S 3 that the theory lives on). Gauge Theory and Strings the supergravity approximation. z → λz.18) Thus. In the gravity solution (2. (2. What makes this case diﬀerent is that we have the additional parameter R.15) by eliminating λ through the change of variables x → λx. This can be seen from (2. for δ 1.16) corresponds to a UV cutoﬀ in the ﬁeld theory at distances δ. So. we see that the AdS/CFT correspondence saturates the holographic bound [66]. From the form of the full metric on AdS5 × S 5 we see that as we take δ → 0 the physical size of four of the dimensions of this nine dimensional space grow.

As discussed in Section 2. In Section 2 we saw how string theory on AdS deﬁnes a partition function which can be used to deﬁne a ﬁeld theory. the AdS/CFT correspondence is a strong/weak coupling duality. 1 Unlike most of the other tests described here. we can compute the equation of state of the SYM theory and also the quark-antiquark potential both for small λ and for large λ. As described in Section 2. since (with our current knowledge) we can only compute most of √ them perturbatively in λ on the ﬁeld theory side and perturbatively in 1/ λ on the string theory side. it relates 2 the region of weak ﬁeld theory coupling λ = gYM N in the SYM theory to the region of high curvature (in string units) in the string theory. Additional ZN symmetries arise when the theories are compactiﬁed on non-simply-connected manifolds. most of the tests described here can be generalized also to cases in other dimensions and/or with less supersymmetry. which cannot change as we change the coupling (except for extreme values of the coupling). Also. Maldacena: Lectures on Large N Field Theories and Gravity 3 Tests of the AdS/CFT correspondence 341 In this section we review the direct tests of the AdS/CFT correspondence. For example. These are: • the global symmetries of the theory. so they can be compared to test the duality. We will focus here only on tests of the correspondence between the N = 4 SU (N ) SYM theory and the type IIB string theory compactiﬁed on AdS5 × S 5 . which will be described below. Thus. and vice versa. Z) duality of the N = 4 SYM theory itself).J. In the ’t Hooft large N limit. in the case of the AdS/CFT correspondence we have the same supergroup SU (2. as described below. 2|4) (whose bosonic subgroup is SO(4. Here we will review the evidence showing that this ﬁeld theory is indeed the same as the conjectured dual ﬁeld theory. a direct comparison of correlation functions is generally not possible. which we do not know how to compare since we can only compute them perturbatively on both sides. and we obtain diﬀerent answers. These are the only symmetries of the theory on R4 . and it was realized that there are several properties of these theories which do not depend on the coupling. and these were also successfully matched in [40. Z) duality symmetry acting on their coupling constant τ . A similar situation arises also in many ﬁeld theory dualities that were analyzed in the last few years (such as the electric/magnetic SL(2. 67]1. 2) × SU (4)) as the global symmetry of both theories. . this test actually tests the ﬁnite N duality and not just the large N limit. both theories are believed to have a non-perturbative SL(2.

69]. QβB } = 0.342 Unity from Duality: Gravity. Gauge Theory and Strings • some correlation functions. where α is a Weyl-spinor index (in the 2 of the SO(3. 1) Lorentz group) and A is an index in the 4 of the SU (4)R R-symmetry group (lower indices A will be taken to transform in the ¯ 4 representation). but it is not clear how to see that this is the exact moduli space on the string theory side (especially since high curvatures arise for generic points in the moduli space). In the SU (N ) ﬁeld theory the moduli space is R6(N −1) /SN .1 below. and in the absence of phase transitions). .2 below. • the moduli space of the theory also does not depend on the coupling. which are usually related to anomalies. On the AdS side it is not clear exactly how to deﬁne the moduli space. parametrized by the eigenvalues of six commuting traceless N × N matrices. QB } = {QαA . α β (3.1 The spectrum of chiral primary operators 3. We will also not discuss here tests involving the behavior of the theory on its moduli space [60. which we will not discuss in this section. • the spectrum of chiral operators does not change as the coupling varies. such as the existence of conﬁnement for the ﬁnite temperature theory [68]. 3) are the Pauli matrices and (σ 0 )αα = −δαα (we use the ˙ ˙ conventions of Wess and Bagger [70]).1 The ﬁeld theory spectrum The N = 4 supersymmetry algebra in d = 4 has four generators QA (and α ¯˙ their complex conjugates QαA ). There are however multicenter solutions so that one might think that there is a background of string theory corresponding to any point in the ﬁeld theory moduli space. ˙ α ¯˙ ¯˙ {QA . and it will be compared in Section 3. There are many more qualitative tests of the correspondence. The matching of these correlation functions will be described in Section 3. QαB } = 2(σ µ )αα Pµ δB . They obey the algebra A ¯˙ {QA . 61.1) where σ i (i = 1. • the qualitative behavior of the theory upon deformations by relevant or marginal operators also does not depend on the coupling (at least for chiral operators whose dimension does not depend on the coupling. 3.1. are protected from any quantum corrections and do not depend on λ. 2.

and its ﬁeld content is a vector multiplet in the adjoint of the gauge group. It is natural to divide the operators into single-trace operators and multiple-trace operators.J. The interactions of the theory include a scalar potential proportional to I.2) .2). Dν ]. φ ]. such that the moduli space of the theory is the space of commuting matrices φI (I = 1. In the ’t Hooft large N limit correlation functions involving multiple-trace operators are suppressed by powers of N compared to those of single-trace operators involving the same ﬁelds. Dµ is the covariant derivative.J Tr([φI . or to an N = 1 theory with three chiral multiplets Φi in the adjoint representation (in the 32/3 of the SU (3) × U (1)R ⊂ SU (4)R which is left unbroken by the choice of a single N = 1 SUSY generator) and a superpotential of the form W ∝ ijk Tr(Φi Φj Φk ). The spectrum of operators in this theory includes all the gauge invariant quantities that can be formed from the ﬁelds described above. Maldacena: Lectures on Large N Field Theories and Gravity 343 N = 4 supersymmetry in four dimensions has a unique multiplet which does not include spins greater than one. φJ ]2 ). We will discuss here in detail only the single-trace operators. where σ µν are the generators of the Lorentz group in the spinor representation. λβB } ∼ (σ µν )αβ Fµν + α ¯˙ {QA . 4. φI ] ∼ λαB . For the SU (N ) theory described above. Such a ﬁeld theory is equivalent to an N = 2 theory with one hypermultiplet in the adjoint representation. 4→ An N = 4 supersymmetric ﬁeld theory is uniquely determined by specifying the gauge group. 1) Lorentz group). The classical action of the supersymmetry generators on these ﬁelds is schematically given (for onshell ﬁelds) by [QA . 6). α {QA . the ﬁeld strength Fµν ≡ [Dµ . which is the vector multiplet. · · · . ¯ with similar expressions for the action of the Q’s. In this section we will focus on local operators which involve ﬁelds taken at the same point in space-time. . It includes a vector ﬁeld Aµ (µ is a vector index of the SO(3. and six real scalars φI (where I is an index in the 6 of SU (4)R ). Aµ ] ∼ (σµ )αα λA ˙ β α αβ ˙ ˙ I J αβ [φ . and we have suppressed the SU (4) Clebsch-Gordan coeﬃcients corresponding to the products 4 × 6 → ¯ 4× ¯ 1+15 and 4×4 → 6 in the ﬁrst three lines of (3. properties of the adjoint representation of SU (N ) determine that such operators necessarily involve a product of traces of products of ﬁelds (or the sum of such products). (3. ˙ α β ¯˙ [QA . 4 four complex Weyl fermions λαA (in the ¯ of SU (4)R ). the multiple-trace operators appear in operator product expansions of products of single-trace operators. λB } ∼ (σ µ )αβ Dµ φI .

For the N = 4 supersymmetry algebra (not including the conformal algebra) it is known that medium representations. which are in short representations of the superconformal algebra and are annihilated by some of the supercharges. Gauge Theory and Strings It is natural to classify the operators in a conformal theory into primary operators and their descendants. and the range of helicities is between λ − 2 and λ + 2. which are descendants of the conformal algebra). All other local gauge invariant operators in the theory involve derivatives or products of these operators. acting with additional supercharges always leads to descendants of the conformal algebra (i. More details on the structure of representations of the N = 4 superconformal algebra may be found in [79–85] and references therein. . This representation is usually called the doubleton representation. generated by acting on the lowest state with products of diﬀerent supercharges. and it does not appear in the SU (N ) SYM theory (though the representations which do appear can all be formed by tensor products of the doubleton representation). for instance. the only gaugeinvariant “single trace” operators are the ﬁelds of the vector multiplet itself ¯ (which are φI . These operators form an ultrashort representation of the N = 4 algebra whose range of helicities is from (−1) to 1 (acting with more than two supercharges of the same helicity on any of these states gives either zero or derivatives. which is annihilated by the operators S and Kµ .344 Unity from Duality: Gravity. A generic primary representation of the superconformal algebra will thus include 216 primaries of the conformal algebra. and non-chiral primary operators.e. whose range of helicities is 6. the primary ﬁelds in such representations will have a range of helicities between λ − 4 (if the lowest dimension operator ψ has helicity λ) and λ + 4 (acting with more than 8 supercharges of the same helicity either annihilates the state or leads to a conformal descendant). on the moduli space of the SU (N ) N = 4 SYM theory [71–78]). it is not clear if such medium representations of the superconformal algebra [79] can appear in physical theories or not (there are no known examples). In non-generic representations of the superconformal algebra a product of less than 16 diﬀerent Q’s annihilates the lowest dimension operator. The N = 4 supersymmetry algebra involves 16 real supercharges. In particular. Since the supercharges have helicities ±1/2. and the range of helicities appearing is smaller. Representations of the superconformal algebra are formed by starting with some state of lowest dimension. In a superconformal theory it is also natural to distinguish between chiral primary operators. λA and Fµν = ∂[µ Aν] ). derivatives). in the small representations of the N = 4 superconformal algebra only up to 4 Q’s of the same helicity acting on the lowest dimension operator give a non-zero result. In the U (1) N = 4 SYM theory (which is a free theory). can also exist (they arise. λA . and acting on it with the operators Q and Pµ .

Another way to check this is to see if by acting with Q’s on these operators we get the most general possible states or not. so one would expect the lowest components of the chiral primary representations to be built only from the scalar ﬁelds. 0) of SU (4)R . The lowest component of a superconformal-primary multiplet is characterized by the fact that it cannot be written as a supercharge Q acting on any other operator. Next. though in principle there could be representations where “null vectors” appear only at higher levels). namely if the representation contains “null vectors” or not (it turns out that in all the relevant cases “null vectors” appear already at the ﬁrst level by acting with a single Q.2) it is easy to see that for symmetric traceless representations we get “null vectors” while for other representations we do not. let us analyze in detail the case n = 2. There is no known simple systematic way to compute the full spectrum of chiral primary operators of the N = 4 SU (N ) SYM theory. This is a representation of weight (0. so we see that if some of the indices are antisymmetric the ﬁeld will be a descendant. One possibility is to use the relation between the dimension of chiral primary operators and their R-symmetry representation [96–100]. ψ} for any ﬁeld ψ. The symmetric product of two 6’s is given by 6 × 6 → 1 + 20 . Let us analyze operators of the form OI1 I2 ···In ≡ Tr(φI1 φI2 · · · φIn ). Maldacena: Lectures on Large N Field Theories and Gravity 345 In the context of AdS space one can think of this multiplet as living purely on the boundary of the space [46. as expected for the U (1) part of the original U (N ) gauge group of the D3-branes (see the discussion in Sect. and to check if this relation is obeyed in the free ﬁeld theory. The ﬁeld in the 1 representation is Tr(φI φI ). so we will settle for presenting the known chiral primary operators. There are several diﬀerent ways to answer this question. Tr(φI φI )] ∼ C AJB Tr(λαB φJ ) where C AIB is a Clebsch-Gordan coeﬃα cient for ¯ × 6 → 4. Looking at the action of the supersymmetry charges (3. In this way we ﬁnd that the representation is chiral primary if and only if the indices form a symmetric traceless product of n 6’s (traceless representations are deﬁned as those who give zero when any two indices are contracted). 2). and this turns out to be correct. First we can ask if this operator can be written as {Q. only symmetric combinations of the indices will be lowest components of primary multiplets. for which [QA .86–95]. For instance. In the SUSY algebra (3. Using the SUSY algebra (3.J. we should ask if the multiplet built on such an operator is a (short) chiral primary multiplet or not. which is 4 .2) only commutators of φI ’s appear on the right hand side. The right-hand side is in the 4 representation. n. where [OI1 I2 ···In ] = n.2) suggests that generally operators built from the fermions and the gauge ﬁelds will be descendants (given by Q acting on some other ﬁelds). Thus. in this section we will refer to SU (4)R representations either by their dimensions in boldface or by their weights.

3. The last issue we should discuss is what is the range of values of n. 84. 101] for more details) one can verify that chiral primary representations correspond exactly to symmetric traceless products of 6’s. up to operators including commutators which (as explained above) are descendants of the SUSY algebra. 81. If we use an N = 1 subalgebra of the N = 4 algebra. The full chiral primary representations are obtained by acting on the ﬁelds On by the generators Q and P of the supersymmetry algebra. the operators On include the chiral operators of the form Tr(Φi1 Φi2 · · · Φin ) (in a representation of SU (3) which is a symmetric product of 3’s). . Tr(φ{I φJ} )} ∼ Tr(λαB φK ) with the rightα hand side being in the 20 representation (appearing in ¯ × 6 → 4 + 20).346 Unity from Duality: Gravity. so it does not form an independent operator. Since the 60 does not appear on the right-hand side (it is a “null vector”) we identify that the representation built on the 20 is a short representation of the SUSY algebra. This means that for n > N we can express the operator OI1 I2 ···In in terms of other operators. N . we ﬁnd that the short chiral primary representations are built on the operators On = O{I1 I2 ···In } with n = 2. The 1 representation built on On contains a total of 256 × 12 n2 (n2 − 1) primary states. The superconformal algebra determines the dimension of these ﬁelds to be [On ] = n. we ﬁnd that {QA . We argued above that these are the only short chiral primary representations in the SU (N ) gauge theory. Thus. Gauge Theory and Strings the most general representation that can appear in the product 4 × 1. for which the indices are in the symmetric traceless product of n 6’s (in a U (N ) theory we would ﬁnd the same spectrum with the additional representation corresponding to n = 1). On the other hand. · · · . By similar manipulations (see [24. as discussed above. but for a particular choice of the N = 1 subalgebra not all the operators On appear to be chiral (a short multiplet of the N = 4 algebra includes both short and long multiplets of the N = 1 subalgebra). It is possible to analyze the chiral primary spectrum also by using N = 1 subalgebras of the N = 4 algebra. commutators always lead to descendants. as described above. of which half are bosonic and half are fermionic. The product of more than N commuting2 N × N matrices can always be written as a sum of products of traces of less than N of the matrices. which is the same as their value in the free ﬁeld theory. Since these 2 We can limit the discussion to commuting matrices since. 4 while the left-hand side could in principle be in the 4 × 20 → 20 + 60. but we will not attempt to rigorously prove this here. so we ﬁnd no null vectors at this level. if we look at the symmetric traceless product Tr(φ{I φJ} ) ≡ Tr(φI φJ )− 1 δ IJ Tr(φK φK ) in the 20 6 representation. and we can write any product of matrices as a product of commuting matrices plus terms with commutators.

in the (0. Both of these ﬁelds are complex. There is an elegant way to write these multiplets as traces of products of “twisted chiral N = 4 superﬁelds” [81. 0) representation of SU (4)R . In the ﬁrst case we get a Lorentz scalar ﬁeld in the (2. and its dimension is n+4. see also [102] which checks some components of these superﬁelds against the couplings to supergravity modes predicted on the basis of the DBI action for D3-branes in anti-de Sitter space [4]. [Qβ . with the complex conjugate ﬁelds given by ¯ ¯ the action of two Q’s. The highest dimension primary ﬁeld ¯ in the multiplet is (generically) of the form Q4 Q4 On . 0) representation. which is of the schematic form αβ {Qα . n − 4. φK2 ]φL1 · · · φLn−1 ). 101]. their Lorentz indices α can be either antisymmetrized or symmetrized. ˙ α ˙ At dimension n + 2 (acting with four supercharges) we ﬁnd: • a complex scalar ﬁeld in the (0. Using an N = 1 subalgebra some of these operators may be written as 2 the lowest components of the chiral superﬁelds Tr(Wα Φj1 · · · Φjn−2 ). Maldacena: Lectures on Large N Field Theories and Gravity 347 multiplets are built on a ﬁeld of helicity zero. For example. In the second case we get an anti-symmetric 2-form of the Lorentz group. n−2.2). ¯ • a real scalar ﬁeld in the (2. B1 ¯ B2 I1 αβ αβ ˙ ˙ In−4 ¯ Tr(λαA1 λβA2 λα λβ φ · · · φ ) + ···. of the form {Q{α . given by Q2 Q2 On . n − 2. If we 2 act with two supercharges QA of the same chirality. On ]} ∼ Tr((σ µν )αβ Fµν φJ1 · · · φJn−1 ) + Tr(λαA λβB φK1 · · · φKn−2 ). It is easy to ﬁnd the form of all the ﬁelds in such a multiplet by using the algebra (3. n − 2. they will contain primary ﬁelds of helicities between (−2) and 2. 1) representation of SU (4)R . let us analyze here in detail the bosonic primary ﬁelds of dimension n + 1 in the multiplet. of the form ˙ ˙ .J. of the form ¯˙ {Qα . On ]} ∼ ¯ Tr(λαA λB φJ1 · · · φJn−2 ) + (σ µ )αα Tr((Dµ φJ )φK1 · · · φKn−1 ). [Qα . [Qβ} . To get a ﬁeld of dimension n + 1 we need to act on On with two supercharges (recall that [Q] = 1 ). 0) representation of SU (4)R . given by Q4 On . 2 of the form Tr(Fµν φI1 · · · φIn−2 ) + · · · . On ]} ∼ αβ Tr(λαA λβB φJ1 · · · φJn−2 ) + Tr([φK1 . Acting with one Q and one Q on the state On gives a (real) Lorentz-vector ﬁeld in the (1. 2) representation. n − 1.

The n = 2 multiplet also includes a complex scalar ﬁeld which is an SU (4)R -singlet. 0) representation. of the form Tr(Fµν Dν φJ φI1 · · · φIn−2 ) + · · · . Single-trace . given by Q4 Q4 On . Thus. ﬁelds on AdS5 are in a one-to-one correspondence with operators in the dual conformal ﬁeld theory. Gauge Theory and Strings • a complex vector ﬁeld in the (1. of the form Tr(D{µ φJ Dν} φK φI1 · · · φIn−2 ) + · · · The spectrum of primary ﬁelds at dimension n + 3 is similar to that of dimension n + 1 (the same ﬁelds appear but in smaller SU (4)R representations).2 The string theory spectrum and the matching As discussed in Section 2. The n = 2 representation includes the currents of the superconformal algebra. of the form 4 I1 In−4 ) + · · · . given by ¯ Q2 Q2 On . • an complex anti-symmetric 2-form ﬁeld in the (2. 3 the short multiplets are even shorter since some of the representations appearing above vanish. Note that ﬁelds with more than four Fµν ’s or Tr(Fµν φ · · · φ more than eight λ’s are always descendants or non-chiral primaries. and a symmetric tensor ﬁeld of dimension 4 which is the energy-momentum tensor (the other currents of the superconformal algebra are descendants of these). and a complex scalar in the 1 of dimension 4). φJ2 ]φI1 · · · φIn−2 )+· · · . n − 2. the n = 3 multiplet (which contains a scalar in the 50 of dimension 3 and a complex scalar in the 45 of dimension 4). of the form Tr(Fµν [φJ1 . Fields on AdS naturally lie in the same multiplets of the conformal group as primary operators. and the n = 4 multiplet which contains a scalar in the 105 of dimension 4.1. a complex scalar in the 10 of dimension 3. the second Casimir of these representations is C2 = ∆(∆ − 4) for a primary scalar ﬁeld of dimension ∆ in the ﬁeld theory. whose real 2 part is the Lagrangian density coupling to 4g1 (of the form Tr(Fµν ) + · · · ) 2 YM and whose imaginary part is the Lagrangian density coupling to θ (of the form Tr(F ∧ F )). and C2 = m2 R2 for a ﬁeld of mass m on an AdS5 space with a radius of curvature R. given by Q2 Q2 On . and at dimension n + 4 there is a single primary ﬁeld. • a symmetric tensor ﬁeld in the (0. given by ¯ Q3 QOn .1. 0) representa¯ tion. n − 4. for n = 2 the highestdimension primaries in the chiral primary multiplet have dimension n + 2 = 4. n − 4. 1) representation.348 Unity from Duality: Gravity. 0) representation. which is a ¯ real scalar in the (0. the spectrum of operators described in Section 3.1. In particular. For n = 2. n − 3. It includes a vector of dimension 3 in the 15 representation which is the SU (4)R R-symmetry current. 3.1 should agree with the spectrum of ﬁelds of type IIB string theory on AdS5 × S 5 . For later use we note that the chiral primary multiplets which contain scalars of dimension ∆ ≤ 4 are the n = 2 multiplet (which has a scalar in the 20 of dimension 2.

for I = 1. The computation involves expanding the ten dimensional ﬁelds in appropriate spherical harmonics on S 5 . These ﬁelds all have helicities between (−2) and 2. ∞. These spherical harmonics are in representations corresponding to symmetric traceless products of 6’s of SU (4)R . · · · . d are indices on S 5 . Presumably none of these states are in small multiplets of the superconformal algebra (at least. and the Y k are the scalar spherical harmonics on S 5 . so it is clear that they all lie in small multiplets of the superconformal algebra. which are built on a lowest dimension ﬁeld which is a scalar in the (0. n. For example. . and the equations of motion determine the mass of this ﬁeld to be m2 = k(k + 4)/R2 . gs N . 3. 6 and with 6 I 2 5 k I=1 (y ) = 1. it is not known how to compute the full spectrum of type IIB string theory on AdS5 ×S 5 . 0) representation of SU (4)R for n = 2. y) = k=0 τ k (x)Y k (y) where x is a coordinate on AdS5 . while multiple-trace operators correspond to multi-particle states. are coordinates on S . we ﬁnd a ﬁeld τ (x) on AdS5 in each such (0. and diagonalizing the equations to give equations of motion for free (massless or massive) ﬁelds3 . c. b. plugging them into the supergravity equations of motion. with masses of the order of the string scale 1/ls or of the Planck scale 1/lp . In fact. we ﬁnd representations of the form described in the previous section. where a. they may be written as Y k (y) ∼ y I1 y I2 · · · y Ik where the y I . A similar k expansion may be performed for all other ﬁelds. this would be the prediction of the AdS/CFT correspondence). the only known states are the states which arise from the dimensional reduction of the ten-dimensional type IIB supergravity multiplet.J. The spectrum of type IIB supergravity compactiﬁed on AdS5 × S 5 was computed in [103]. the ten dimensional dilaton ∞ ﬁeld τ may be expanded as τ (x. · · · . The scalar ﬁelds of dimension 3 The ﬁelds arising from diﬀerent spherical harmonics are related by a “spectrum generating algebra”. Maldacena: Lectures on Large N Field Theories and Gravity 349 operators in the ﬁeld theory may be identiﬁed with single-particle states in AdS5 . String theory on AdS5 × S 5 is expected to have many additional states. Thus. y is a coordinate on S 5 . see [104]. k. linearized around the AdS5 ×S 5 background. and we will describe below how they match with the small multiplets of the ﬁeld theory described above. 0) representation of SU (4)R . 2. Such states would correspond (using the mass/dimension relation described above) to operators in the ﬁeld theory with dimensions of order ∆ ∼ (gs N )1/4 or ∆ ∼ N 1/4 for large N. The lowest dimension scalar ﬁeld in each representation turns out to arise from a linear combination of spherical harmonic modes of the S 5 components of the graviton ha (expanded around the AdS5 × S 5 vacuum) and the 4-form a ﬁeld Dabcd . If we organize the results of [103] into representations of the superconformal algebra [80]. Unfortunately.

but this will not aﬀect most of our discussion in this review so we will ignore this possibility here. the n = 2 representation is called the supergraviton representation. the n = 1 doubleton representation. so the results agree within their range of validity.350 Unity from Duality: Gravity. In some generalizations of the AdS/CFT correspondence the string coupling goes to zero at the boundary even for ﬁnite N . N . Gauge Theory and Strings n + 1 correspond to 2-form ﬁelds Bab with indices in the S 5 . and they may be set to zero there. Comparing the results described above with the results of Section 3. and massless SU (4)R gauge ﬁelds which correspond to (or couple to) the global SU (4)R currents in the ﬁeld theory. and then classical string theory should lead to exactly the same . The ﬁeld/operator correspondence matches this representation to the representation including the superconformal currents in the ﬁeld theory. This is one of the reasons why it seems more natural to view the corresponding gauge theory as an SU (N ) gauge theory and not a U (N ) theory. 3. corresponding to a free U (1) vector multiplet in the dual theory.1. In particular. the modes of this multiplet are all pure gauge modes in the bulk of AdS5 . The dilaton ﬁelds described above are the complex scalar ﬁelds arising with dimension n + 2 in the multiplet (as described in the previous subsection). N = 8 gauged supergravity. which (as expected) corresponds to the energy-momentum tensor in the ﬁeld theory. In the naive dimensional reduction of the type IIB supergravity ﬁelds. also appears. and that there are no chiral ﬁelds above this scale. and it includes the ﬁeld content of d = 5. we see that we ﬁnd the same spectrum of chiral primary operators for n = 2. It is not known how to check this prediction. · · · . It includes a massless graviton ﬁeld. Thus. but instead it is equivalent to a topological theory in the bulk) to obtain a theory which is dual to the U (N ) gauge theory.1. tree-level string theory is certainly not enough for this since when gs = 0 we must take N = ∞ to obtain a ﬁnite value of gs N . The ﬁeld theory results suggest that the exact spectrum of chiral representations in type IIB string theory on AdS5 × S 5 actually matches the naive supergravity spectrum up to a mass scale m2 ∼ N 2 /R2 ∼ 2 N 3/2 Mp which is much higher than the string scale and the Planck scale. However. with our current knowledge the matching of chiral primaries of the N = 4 SYM theory with those of string theory on AdS5 × S 5 tests the duality only in the large N limit. The symmetric tensor ﬁelds arise from the expansion of the AdS5 -components of the graviton. It may be possible (and perhaps even natural) to add the doubleton representation to the theory (even though it does not include modes which propagate in the bulk of AdS5 . The supergravity results cannot be trusted for masses above the order of the string scale (which corresponds to n ∼ (gs N )1/4 ) or the Planck scale (which corresponds to n ∼ N 1/4 ).

grows at 2 least as (gs N )1/4 ∼ (gYM N )1/4 . In the SU (N ) gauge theory this number is N 2 − 1. (Dµ Jµ )a = N 2 − 1 abc id 384π 2 µνρσ b c Fµν Fρσ . and (unlike many other theories) these symmetries are exact also in the full quantum theory.3) How can we see this eﬀect in string theory on AdS5 ×S 5 ? One way to see it is. when the theory is coupled to external gravitational or SU (4)R gauge ﬁelds. Thus. In other instances of the AdS/CFT correspondence (such as the ones discussed in [106–108]) there exist also additional chiral primary multiplets with n of order N . These currents are chiral since the fermions λαA are in ¯ the ¯ representation while the fermions of the opposite chirality λA are in 4 α ˙ the 4 representation. it is a prediction of the AdS/CFT correspondence. This happens in particular for the near-horizon limit of NS5-branes. However. (3. Maldacena: Lectures on Large N Field Theories and Gravity 351 spectrum of chiral operators as the ﬁeld theory. 3. in which case the exact spectrum was successfully compared in [105]. if we gauge the SU (4)R global symmetry. The anomaly can be expressed either in terms of the 3-point function of the SU (4)R global currents. these symmetries are broken by quantum eﬀects. T c }) and we take only the negative parity component of the correlator. the dimension of all non-chiral operators in the ﬁeld theory.2 Matching of correlation functions and anomalies The classical N = 4 theory has a scale invariance symmetry and an SU (4)R R-symmetry. In ﬁeld theory this breaking comes from one-loop diagrams and does not receive any further corrections. and these have been successfully matched with wrapped branes on the string theory side. The fact that there seem to be no non-chiral ﬁelds on AdS5 with a mass below the string scale suggests that for large N and large gs N . a b c Jµ (x)Jν (y)Jρ (z) − =− N 2 − 1 abc Tr [γ5 γµ ( x− y)γν ( y− z)γρ ( z− x)] id . of course. to use the general prescription of Section 4 to compute the 3-point . which is proportional to the number of charged fermions. thus it can be computed also in the strong coupling regime and compared with the results from string theory on AdS space. The reason for this behavior on the ﬁeld theory side is not clear. We will begin by discussing the anomaly associated with the SU (4)R global currents. or in terms of the non-conservation of the SU (4)R a current when the theory is coupled to external SU (4)R gauge ﬁelds Fµν . such as Tr(φI φI ). 32π 6 (x − y)4 (y − z)4 (z − x)4 where dabc = 2Tr(T a {T b .J. we will ﬁnd an Adler-Bell-Jackiw anomaly from the triangle diagram of three SU (4)R currents.

7) . the action will change by a boundary term of the form − iN 2 384π 2 d4 x ∂AdS5 µνρσ abc d b c Λa Fµν Fρσ . µ λ σ (3.5) From this we can read oﬀ the anomaly in (Dµ Jµ ) since. (3. to consider directly the meaning of the anomaly (3.3) from the point of view of the AdS theory [24]. It is more illuminating. In the AdS theory we have gauge ﬁelds Aa which couple. but the anomaly means that when we turn on non-zero ﬁeld strengths for these ﬁelds the theory should no longer be gauge invariant. Gauge Theory and Strings function (3. which is of the form iN 2 96π 2 d5 x(dabc AdS5 µνλρσ Aa ∂ν Ab ∂ρ Ac + · · · ). This eﬀect is precisely reproduced by a Chern-Simons term which exists in the low-energy supergravity theory arising from the compactiﬁcation of type IIB supergravity on AdS5 × S 5 . one recovers the term proportional to N 2 ). which we will not discuss here).352 Unity from Duality: Gravity.4) This term is gauge invariant up to total derivatives. where E4 = 1 (R2 − 4R2 + R2 ). µν 16π 2 µνρσ 1 1 2 2 I4 = − R2 . 110] the correct answer to leading order in the large N limit (namely. as explained above. µνρσ − 2Rµν + R 2 16π 3 (3. The other anomaly in the N = 4 SYM theory is the conformal (or Weyl) anomaly (see [111. In four dimensions.112] and references therein). the general form of the conformal anomaly is g µν Tµν = −aE4 − cI4 .6) (3. when we have a a coupling of the form d4 xAµ Jµ .3) for large N . however. which means that if we take a gauge transformation Aa → Aa + (Dµ Λ)a for which Λ does not µ µ vanish on the boundary of AdS5 .3). and indeed one ﬁnds [109. The conformal anomaly is related to the 2-point and 3-point functions of the energy-momentum tensor [113–116]. the change in the action under a gauge a 4 a a transformation is given by d x(Dµ Λ)a Jµ = − d4 xΛa (Dµ Jµ ). indicating the breakdown of conformal invariance when the theory is coupled to a curved external metric (there is a similar breakdown of conformal invariance when the theory is coupled to external SU (4)R gauge ﬁelds. to µ a the SU (4)R global currents Jµ of the gauge theory. and we ﬁnd exact agreement with (3.

and also all the loop corrections. which are of order N 2 . where the conformal anomaly was shown to arise from subtleties in the regularization of the (divergent) supergravity action on AdS space. if we do not require the string theory to have a supergravity approximation then there is no such restriction. such corrections were discussed in [122] and successfully compared in [123–125]5 . 24].1. Throughout this section.J. 4 Correlation functions A useful statement of the AdS/CFT correspondence is that the partition function of string theory on AdS5 × S 5 should coincide with the partition function of N = 4 super-Yang-Mills theory “on the boundary” of AdS5 [23. we approximate the string theory partition function by e−ISUGRA . they are successful tests of the duality in the large N limit. 5 Computing 4A . For other instances of the AdS/CFT correspondence there are corrections to anomalies at order 1/N ∼ gs (α /R2 )2 . it seems worthwhile to motivate the methodology from a somewhat diﬀerent perspective. A free ﬁeld computation in the SU (N ) µ N = 4 SYM theory leads to a = c = (N 2 −1)/4. generalization with more varying ﬁelds may be found in [121]. and to leading order in the large N limit it should be obtained from type IIB supergravity on AdS5 × S 5 . the same result should be obtained in type IIB string theory on AdS5 × S 5 . The result of [117–120] implies that a computation using gravity on AdS5 always gives rise to theories with a = c. this is weaker than the statement that the correspondence holds for ﬁnite N . This was indeed found to be true in [117–120]4 . Of course. however. Maldacena: Lectures on Large N Field Theories and Gravity 353 where Rµνρσ is the curvature tensor. so generalizations of the AdS/CFT correspondence which have (for large N ) a supergravity approximation are limited to conformal theories which have a = c in the large N limit. In supersymmetric theories the supersymmetry algebra relates g µν Tµν to derivatives of the R-symmetry current. Thus. but before summarizing the actual calculations of Green’s functions. µρν and R ≡ Rµ is the scalar curvature. so it is protected from any quantum corrections. It would be interesting to compare other corrections to the large N result. since the 1/N expansion is not expected to converge. Rµν ≡ Rρ is the Riemann tensor. such corrections tests the conjecture that the correspondence holds order by order in 1/N . Thus. which are controlled essentially by the gravitational coupling κ ∼ gst α 2 . For both of the anomalies we described the ﬁeld theory and string theory computations agree for the leading terms. The basic idea was explained in Section 2. On the gauge theory side. where ISUGRA is the supergravity action evaluated on AdS5 × S 5 (or on small deformations of this space). This approximation amounts to ignoring all the stringy α corrections that cure the divergences of supergravity.

2) we have not attempted to be prescient about inserting factors of .2) without modiﬁcation to compute two-point functions of O.1) where W is the generating functional for connected Green’s functions in the gauge theory. φ z= (4. we arrive at the relation ISUGRA βF . we think of closed string states in the bulk as sourcing gauge singlet operators on the brane which originate as composite operators built from open strings. we think of choosing the values φ0 arbitrarily and then extremizing the action ISUGRA [φ] in the region z > subject to these boundary conditions. so the main claim that is being made here is that the dual gauge theory provides a description of the state of the black hole which is physically equivalent to the one in string theory.1. If there is more than one solution. The true boundary of anti-de Sitter space is z = 0. is the on-shell supergravity action. this approximation amounts to taking both N 2 and gYM N large. gYM N limit. Instead our strategy will be to use (4. We will discuss the ﬁnite temperature case further in Section 6. In the supergravity approximation. At ﬁnite temperature. When we apply this relation to a Schwarzschild black hole in AdS5 . Gauge Theory and Strings as explained in Section 2. and the basic relation becomes e−ISUGRA Zstring = Zgauge = e−W . with value φ0 (x) for z = . we can write Wgauge [φ0 ] = − log e Ê d4 x φ0 (x)O(x) CFT extremum ISUGRA [φ] . The main technical idea behind the bulk-boundary correspondence is that the boundary values of string theory ﬁelds (in particular. Note that in (4. We will write the bulk ﬁelds generically as φ(x. multiple saddle points will not be a problem. which is thought to be reﬂected in the gauge theory by a thermal state at the Hawking temperature of the black hole. and then perform a wave-function . In short. From a D-brane perspective. then we have more than one saddle point contributing to the string theory partition function.354 Unity from Duality: Gravity. W = βF where β is the inverse temperature and F is the free energy of the gauge theory. z) (in the coordinate system (2.15)). 2 in the large N. we solve the equations of motion in the bulk subject to Dirichlet boundary conditions on the boundary. and evaluate the action on the solution. and we must determine which is most important.2) =φ0 That is. the generator of connected Green’s functions in the gauge theory. So. and = 0 serves as a cutoﬀ which will eventually be removed. Calculating the free energy of a black hole from the Euclidean supergravity action has a long tradition in the supergravity literature [126]. and devote the rest of this section to the partition function of the ﬁeld theory on R4 . In this section. (4. supergravity ﬁelds) act as sources for gauge-invariant operators in the ﬁeld theory.

In absorption calculations one is keeping the whole D3-brane geometry. 127–141]–have undertaken the program of extracting explicit n-point correlation functions of gauge singlet operators by developing both sides of (4. They are special limits of the usual gravity propagators in the bulk. corresponding to a ﬁeld theory which does not have a conformal ﬁxed point in the ultraviolet. For massive scalar ﬁelds in AdS5 . A remark is in order regarding the relation of (4. the value of the supergravity ﬁeld which the D3-branes feel is the same as the value in the curved space description at the boundary of the near horizon region. not just the near-horizon AdS5 × S 5 throat.2) is just a mathematical expression of the fact that the throat geometry should respond identically to the perturbed supergravity ﬁelds as the low-energy theory on the D3-branes. At small ’t Hooft coupling. Following [23. The usual treatment is to split the space into a near region (the throat) and a far region.J. As usual. In the low-energy limit. The incoming wave from asymptotically ﬂat inﬁnity can be regarded as ﬁxing the value of a supergravity ﬁeld at the outer boundary of the near region. the supergravity description is valid at large N and large ’t Hooft coupling. There is one diﬀerence: in ordinary Feynman graphs one assigns the wavefunctions of asymptotic states to the external legs of the graph. This approach should be workable even in a space (with boundary) which is not asymptotically anti-de Sitter. 2 2 (4. the power series has a graphical representation in terms of treelevel Feynman graphs for ﬁelds in the supergravity. 24]. Because the right hand side is the extremization of a classical action.3) where η is some normalization which in principle follows from the tendimensional origin of the action. and the incoming wave impinges upon it. and are called bulk-to-boundary propagators. but in the present case the external leg factors reﬂect the boundary values φ0 .2) in a power series in φ0 . a number of papers–notably [109. 110. We will encounter their explicit form in the next two sections.2) to the old approach of extracting Green’s functions from an absorption cross-section [16]. this has the generic form S=η 1 √ 1 d5 x g (∂φ)2 + m2 φ2 . Equation (4. only the part of the action which is quadratic in the relevant ﬁeld perturbation is needed. there is a diﬀerent description of the process: a cluster of D3-branes sits at some location in ﬂat ten-dimensional space. Maldacena: Lectures on Large N Field Theories and Gravity 355 renormalization on either O or φ so that the ﬁnal answer is independent of the cutoﬀ. 4.1 Two-point functions For two-point functions. The bulk-to-boundary propagator is a .

z) = φ0 (x) = eip·x at z = . which is useful for computing the two-point function and also in situations where the bulk geometry loses conformal invariance. A complete set of solutions for the linearized equation. ( −m2 )φ = 0. ( −m2 )φ = 0. which corresponds to the Poincar´ coordinates of AdS e space. where Iν and Kν are Bessel functions and ∆=2+ 4 + m2 R 2 . We will always work in Euclidean space6 .7) 6 The results may be analytically continued to give the correlation functions of the ﬁeld theory on Minkowskian Ê4 . To avoid divergences associated with the small z region of integration in (4.4) provides manifest Euclidean symmetry on the directions parametrized by x. namely Z(u) = u2 I∆−2 (u) and Z(u) = u2 K∆−2 (u).5) There are two independent solutions to (4. which has special asymptotic properties. Gauge Theory and Strings particular solution of the equation of motion. z) = (pz)2 K∆−2 (pz) ip·x e .6) The second solution is selected by the requirement of regularity in the interior: I∆−2 (u) increases exponentially as u → ∞ and does not lead to a ﬁnite action conﬁguration. then.3). z ≥ . we ﬁnd the bulk-to-boundary propagator φ(x. (p )2 K∆−2 (p ) (4. (4. We will start by considering the momentum space propagator. is given by φ = eip·x Z(pz). where the function Z(u) satisﬁes the radial equation u5 ∂u 1 ∂u − u2 − m2 R2 Z(u) = 0 . u3 (4. which has proven more convenient for the study of higher point correlators in the conformal case.356 Unity from Duality: Gravity. Imposing the boundary condition φ(x. A coordinate system in the bulk of AdS5 such that ds2 = R2 dx2 + dz 2 z2 (4. .5). z) = Kp (x. we will discuss the position space propagator. we will employ an explicit cutoﬀ.

performed the integral and expanded in . The leading analytic terms give rise to contact terms in position space. From here on we will take that convention. precise matchings between the normalizations in ﬁeld theory and in string theory for all the chiral primary operators have not been worked out. in the spirit of diﬀerential regularization [142].8) we have used (4.7). See [15] for an early approach to this problem. O(x)O(y) = η 2∆−8 2∆ 1 − 4 Γ(∆ + 1) · 2 Γ(∆ − 2) |x − y|2∆ ∆ π Several explanatory remarks are in order: • To establish the second equality in (4. the graviton. The result is a pure power law dependence on the separation |x − y|. z = )O(x) to compute the Green’s functions. substituted in (4. The explicit powers of the cutoﬀ in the ﬁnal position space answer can be eliminated by absorbing a factor of ∆−4 into the deﬁnition of O. we write O(p)O(q) = ∂ 2 W φ0 = λ1 eip·x + λ2 eiq·x ∂λ1 ∂λ2 λ1 =λ2 =0 = (leading analytic terms in ( p)2 ) −η 2∆−8 (2∆ − 4) Γ(3 − ∆) 4 δ (p + q) Γ(∆ − 1) p 2 2∆−4 (4. which amounts to inserting a factor of 4−∆ on the right hand side of (4. Only the leading nonanalytic term is essential.8) + (higher order terms in ( p)2 ). as required by conformal invariance.7). Expanding around integer ∆ ≥ 2 one obtains ﬁnite expressions involving log p. but they can be deﬁned for generic complex ∆ by analytic continuation and for positive integer ∆ by expanding around a pole and dropping divergent terms. • We have assumed a coupling d4 x φ(x.J. and their superpartners . In part this is due to the diﬃculty of determining the coupling of bulk ﬁelds to ﬁeld theory operators (or in stringy terms. In fact. Maldacena: Lectures on Large N Field Theories and Gravity 357 To compute a two-point function of the operator O for which φ0 is a source.3). and the higher order terms are unimportant in the limit where we remove the cutoﬀ. • The Fourier transforms used to obtain the last line are singular. For the dilaton. We have given the expression for generic real values of ∆. the coupling of closed string states to composite open string operators on the brane).

large N limit to which supergravity applies. 109. Conventions seem fairly uniform in the literature. and some of them escape the short multiplets and live in long multiplets of the smaller supergroups.358 Unity from Duality: Gravity. 2 . the couplings can be worked out explicitly. The remarkable fact is that all the string theory modes with m ∼ 1/R (which is to say. but the Kaluza-Klein states are roughly as numerous as before. 2 • spinors: ∆ = 1 (d + 2|m|). with √ m ∼ 1/ α . and typically correspond to dimensions 2 which are ﬁnite (in the large gYM N limit) but irrational. For instance. In some of these cases all normalizations have been worked out unambiguously and checked against ﬁeld theory predictions (see for example [23. This is no longer true for theories of reduced supersymmetry: the supergroup gets smaller. 2 • vectors: ∆± = 1 (d ± 2 • p-forms: ∆ = 1 (d ± 2 (d − 2)2 + 4m2 ). 2|4). For N = 4 super-Yang-Mills theory. • The mass-dimension relation (4. The operators in short multiplets have algebraically protected dimensions. • ﬁrst-order (d/2)-forms (d even): ∆ = 1 (d + 2|m|). (d − 2p)2 + 4m2 ). Correlation functions of non-scalar operators have been widely studied following [24]. All other states have a much larger mass. are expected to correspond to operators with dimen2 sion ∆ ∼ (gYM N )1/4 . 134]). and a table of mass-dimension relations in AdSd+1 with unit radius was made in [154] from the various sources cited above (see also [101]): √ • scalars: ∆± = 1 (d ± d2 + 4m2 ). so in this case it should be enough to study the scalars.6) holds even for string states that are not included in the Kaluza-Klein supergravity reduction: the mass and the dimension are just diﬀerent expressions of the second Casimir of SO(4. the literature includes [143–153]. all correlation functions of ﬁelds in chiral multiplets should follow by application of supersymmetries once those of the chiral primary ﬁelds are known. Gauge Theory and Strings (including gauge ﬁelds in AdS5 ). The obvious conclusion is that all operators whose dimensions are not algebraically protected have large dimension in the strong ’t Hooft coupling. They still have a mass on the order of 1/R. 2). It is worthwhile to note however that the mass-dimension formula changes for particles with spin. all closed string states which arise from massless ten dimensional ﬁelds) fall in short multiplets of the supergroup SU (2. In fact the deﬁnition of mass has some conventiondependence. excited string states.

the bulk-to-boundary propagator K∆ (x. the momentum representation seems uniformly more convenient. In practice this ambiguity is usually resolved by appealing to some special algebraic property of the relevant ﬁelds. 359 In the case of ﬁelds with second order lagrangians. but for higher point functions position space is preferred because the full conformal invariance is more obvious. as discovered in [109].J. This is not . Calculating two-point functions directly using the position-space propagators (4.2 Three-point functions Working with bulk-to-boundary propagators in the momentum representation is convenient for two-point functions. but in some cases (notably m2 = 15/4 in AdS5 ) there is a genuine ambiguity. the Fourier transform of (4. The unique regular K∆ solving the equation of motion and satisfying (4.) The boundary behavior of position space bulk-to-boundary propagators is speciﬁed in a slightly more subtle way: following [109] we require K∆ (x. such as transformation under supersymmetry or a global bosonic symmetry. cut oﬀ all bulk integrals at that boundary.9) is K∆ (x. Usually the choice ∆ = ∆+ is clear from the unitarity bound. y) → z 4−∆ δ 4 (x − y) as z → 0. That is what we have done in (4. Thus at any ﬁnite . A correct prescription is to specify boundary conditions at ﬁnite z = . 4. one arrives at a diﬀerent answer.7) (modiﬁed by the factor of 4−∆ as explained after (4. z. and only afterwards take → 0.9). For brevity we will omit a further discussion of higher spins. we have not attempted to pick which of ∆± is the physical dimension.10) At a ﬁxed cutoﬀ.8)). Maldacena: Lectures on Large N Field Theories and Gravity • spin-3/2: ∆ = 1 (d + 2|m|). but cutting the bulk integrals oﬀ again at .9) Here y is the point on the boundary where we insert the operator. for non-conformal examples of the bulk-boundary correspondence. z) is a point in the bulk. and ﬁnally taking the same → 0 answer.10) only approximately coincides with (4. (However. y) = Γ(∆) π 2 Γ(∆ − 2) z z 2 + (x − y)2 ∆ · (4. 2 • massless spin-2: ∆ = d. This apparently innocuous subtlety turned out to be important for two-point functions. y) is a continuous function which approximates 4−∆ δ 4 (x − y) better and better as → 0.8). z. z = . and (x. and instead refer the reader to the (extensive) literature. (4. .

O1 (x1 )O2 (x2 )O3 (x3 ) = −λ = √ d5 x gK∆1 (x. and φ3 . where Jµ is a conserved current. and the cubic vertex to a factor of λ. 2 i i 2 (4. of the form S= √ d5 x g i 1 (∂φi )2 + 1 m2 φ2 + λφ1 φ2 φ3 . That is.360 Unity from Duality: Gravity. x3 ) λa1 .e. then the calculation of O1 O2 O3 reduces.11) and if the φi couple to operators in the ﬁeld theory by interaction terms d4 x φi Oi . |x1 − x2 |∆1 +∆2 −∆3 |x1 − x3 |∆1 +∆3 −∆2 |x2 − x3 |∆2 +∆3 −∆1 (4. The legs correspond to factors of K∆i . Gauge Theory and Strings 1 2 λ 3 Fig. The authors of [109] checked that using the cutoﬀ consistently (i.2). x1 )K∆2 (x. Two-point functions are uniquely diﬃcult because of the poor convergence properties of the integrals over z. If one has a Euclidean bulk action for three scalar ﬁelds φ1 . The position of the vertex is integrated over AdS5 . The Feynman graph for the three-point function as computed in supergravity. to the evaluation of the graph shown in Figure 2. surprising since the z = boundary conditions were not used consistently. with the momentum space propagators) gave two-point functions O(x1 )O(x2 ) a normalization such that Ward identities involving the three-point function O(x1 )O(x2 )Jµ (x3 ) . φ2 . 2. The integrals involved in three-point functions are suﬃciently benign that one can ignore the issue of how to impose the cutoﬀ.12) . were obeyed. via (4. x2 )K∆3 (x.

The result of [132] is that the equations of motion for the scalar modes sI dual to ˜ I OI = CJ1 . φJ ) (4.13) Γ 1 2 (∆1 In principle one could also have couplings of the form φ1 ∂φ2 ∂φ3 . The result [109] is a1 = − Γ 1 2 (∆1 + ∆2 − ∆3 ) Γ 1 (∆1 + ∆3 − ∆2 ) Γ 1 (∆2 + ∆3 − ∆1 ) 2 2 · 2π 4 Γ(∆1 − 2)Γ(∆2 − 2)Γ(∆3 − 2) + ∆2 + ∆3 ) − 2 . but the only way to compute a1 is by performing the integral over x. These operators are the chiral primaries of the gauge theory: all other single trace operators of protected dimension descend from these by commuting with supersymmetry generators. For AdS5 × S 5 . φJ ) in ﬁeld theory. one must instead expand in ﬂuctuations around the background metric and ﬁve-form ﬁeld strength.15) Derivative couplings of the form s∂˜∂˜ are expected a priori to enter ˜ s s into (4.14) and (4.J.J Trφ(J1 . where the parentheses indicate a symmetrized traceless product.. (4. The old literature [103] only dealt with the linearized equations of motion. 159].15) requires some explanation. .. .14) follow from an action of the form S= 4N 2 (2π)5 d5 x √ g I AI (wI )2 sI −(∇˜I )2 − l(l − 4)˜2 s 2 + I1 . The dependence on the xi is dictated by the conformal invariance. This was done for a restricted set of ﬁelds in [132].15). and we are interested only in modes which are scalars in AdS5 . The ﬁelds considered were those dual to operators of the form Trφ(J1 φJ2 . In ﬂat ten-dimensional space these terms can be read oﬀ directly from the original type IIB supergravity papers [158. Only the metric and the ﬁve-form are involved in the dual supergravity ﬁelds. for 3-point functions it is necessary to go to one higher order of perturbation theory. Maldacena: Lectures on Large N Field Theories and Gravity 361 for some constant a1 . . The main technical diﬃculty with three-point functions is that one must ﬁgure out the cubic couplings of supergravity ﬁelds.I3 GI1 I2 I3 wI1 wI2 wI3 sI1 sI2 sI3 · ˜ ˜ ˜ 3 (4. I is an index which runs .I2 . The notation in (4. but an appropriate ﬁeld redeﬁnition eliminates them. . Because of the diﬃculties in writing down a covariant action for type IIB supergravity in ten dimensions (see however [155–157]). This leads only to a modiﬁcation of the constant a1 . it is most straightforward to read oﬀ these “cubic couplings” from quadratic terms in the equations of motion.

We have taken the coupling to ˜ be d4 x sI OI . (4. as calculated using (4. . . To get rid of factors of wI . and the normalization ambiguity is represented by the “leg factors” wI . It is the combination sI = wI sI rather than sI itself which ˜ ˜ has a deﬁnite relation to supergravity ﬁelds. chosen so that CJ1 . · · · . We refer the reader to [132] for explicit expressions for AI and the symmetric tensor GI1 I2 I3 .. The Wilson loop operator W (C) = Tr P exp i C A (5. but such a theorem has not yet been proven (see however comments at the end of Sect.362 Unity from Duality: Gravity. This suggests that there is a non-renormalization theorem for this correlation function. Gauge Theory and Strings over the weight vectors of all possible representations constructed as symmetric traceless products of the 6 of SU (4)R . These are the representations I whose Young diagrams are .12). ∆1 +∆2 −∆3 |x − x |∆1 +∆3 −∆2 |x − x |∆2 +∆3 −∆1 N |x1 − x2 | 1 3 2 3 where we have deﬁned I I I3 C I1 C I2 C I3 = CJ1 ···Ji K1 ···Kj CJ2 ···Ji L1 ···Lk CK1 ···Kj L1 ···Lk . 5 Wilson loops In this section we consider Wilson loop operators in the gauge theory. is OI1 (x1 )OI2 (x2 )OI3 (x3 ) = √ 1 ∆1 ∆2 ∆3 C I1 C I2 C I3 .J = δ IJ .. into which the “leg factors” wI do not enter.1 leads to OI1 (x)OI2 (0) = δ I1 I2 · x2∆1 (4. 3... This is the same strategy as was pursued in comparing matrix models of quantum gravity to Liouville theory. One can choose wI so ˜ ˜ that a two-point function computation along the lines of Section 4. As commented in the previous section.16) With this choice. It is worth emphasizing that the normalization ambiguity in the bulk-boundary coupling is circumvented essentially by considering invariant ratios of three-point functions and two-point functions. the three-point function. we introduce operators OI = wI OI .17) is the same result one obtains from free ﬁeld theory by Wick contracting all the φJ ﬁelds in the three operators..J is a basis transformation I J matrix.J CJ1 ..17) Remarkably. 1 1 (4. CJ1 .1) .2). there is generally a normalization ambiguity on how supergravity ﬁelds couple to operators in the gauge theory.

to having a D3 brane sitting at some radial position U in AdS. Since they are charged they will act as a source for the vector ﬁelds. we expect the Wilson loop to be related to the string running from the quark to the antiquark. we have W ∼ e−T V (L) . A more formal argument for this coupling is that these states are BPS.1 Wilson loops and minimum surfaces In QCD. 5. from the point of view of the U (N ) gauge theory. Finally. These strings will pull the N branes and will cause a deformation of the branes. In order to motivate this prescription let us consider the following situation. transforming in the N of U (N ). which is a superstring which lives in ten dimensions. one can see . and it involves the path-ordered integral of the gauge connection along the contour. which act as a source for the various U (N ) ﬁelds. and at a point on S 5 . For large N and large gYM N . For this purpose we consider a rectangular loop with sides of length T and L in Euclidean space. Then. and which can stretch between two points on the boundary of AdS. This corresponds. So. In order to get a non-dynamical source (an “external quark” with no ﬂuctuations of its own. as discussed in Section 2. The coupling to the scalar ﬁelds can be seen qualitatively by viewing the quarks as strings stretching between the N branes and the single separated brane. 167]. get a mass proportional to U . We expect this string to be analogous to the string in our conﬁguration. Thus. the string should end on the boundary of AdS space. (5. Thus. m = U/2π.J. the AdS/CFT correspondence maps the computation of W in the CFT into a problem of ﬁnding a minimum surface in AdS [166. viewing T as the time direction. The trace is taken over some representation of the gauge group. which means U should also go to inﬁnity. We start with the gauge group U (N + 1). These stretched strings will also act as a source for the scalar ﬁelds. Maldacena: Lectures on Large N Field Theories and Gravity 363 depends on a loop C embedded in four dimensional space. and we break it to U (N ) × U (1) by giving an expectation value to one of the scalars. which will correspond precisely to the Wilson loop operator) we need to take m → ∞. we will discuss here only the case of the fundamental representation (see [165] for a discussion of other representations). and the coupling to the scalar (Higgs) ﬁelds is determined by supersymmetry. it is clear that for large T the expectation value will behave as e−T E where E is the lowest possible energy of the quark-anti-quark conﬁguration. The oﬀ-diagonal states. we can view these states as massive quarks.2) 2 where V (L) is the quark anti-quark potential. From the expectation value of the Wilson loop operator W (C) we can calculate the quark-antiquark potential. which is described by the scalar ﬁelds.

This is the expression when the signature of R4 is Euclidean. is ﬁnite. where |C| is the length of the loop in the ﬁeld theory [166.3) where xµ (τ ) is any parametrization of the loop and θI (I = 1. So. In the supergravity regime. (5. However. the leading contribution to this partition function will come from the area of the string worldsheet. Generalizing the prescription of section 4 for computing correlation functions. In the Minkowski signature case. 6) is a unit vector in R6 (the point on S 5 where the string is sitting). · · · . when gs N is large. A detailed derivation of (5. as it should be since a divergence in the Wilson loop would have implied a mass renormalization of the BPS particle. 167]. The divergence arises from the fact that the string worldsheet is going all the way to the boundary of AdS. with the condition that we have a string worldsheet ending on the loop C. . The apparent discrepancy between the divergence of the area of the minimum surface in AdS and the ﬁniteness of the ﬁeld theory computation can be reconciled by noting that the appropriate action for the string worldsheet is not the area itself but its Legendre transform with respect to the string coordinates corresponding to θI and the 7 The diﬀerence in the factor of i between the Euclidean and the Minkowski cases can √ be traced to the analytic continuation of x2 . The area as deﬁned above is divergent. the perturbative computation in the ﬁeld theory shows that W . as in Figure 3 [166.3) in N = 4 SYM we should consider the string theory partition function on AdS5 × S 5 . the discussion above implies that in order to compute the expectation value of the operator (5.3).364 Unity from Duality: Gravity. This area is measured with the AdS metric. Gauge Theory and Strings this coupling explicitly by writing the full U (N + 1) Lagrangian. for W given by (5. which will imply some particular boundary conditions for the worldsheet fermions at the boundary of AdS. we see that for large r it diverges as r|C|. the phase factor associated to the trajectory of the quark has an extra factor “i” in front of θI 7 . If we evaluate the area up to some radial distance U = r. the ﬁnal conclusion is that the stretched strings couple to the operator W (C) = Tr P exp √ (iAµ xµ + θI φI x2 )dτ ˙ ˙ .3) can be found ˙ in [168]. this will not aﬀect the leading order computations we describe here. 167]. putting in the Higgs expectation value and calculating the equation of motion for the massive ﬁelds [166]. On the other hand. The precise deﬁnition of the Wilson loop operator corresponding to the superstring will actually include also the ﬁeld theory fermions. and it is generally not the same as the area enclosed by the loop C in four dimensions.

4) where e1 . The Wilson loop operator creates a string worldsheet ending on the corresponding loop on the boundary of AdS. As an example let us consider a circular Wilson loop. In the case of the line. using the coordinates (2. We could ﬁnd the surface that minimizes the area by solving the Euler-Lagrange equations. x= a2 − z 2 (e1 cos θ + e2 sin θ). leaving the resulting action ﬁnite. (5. Using the conformal transformation to map the line to a circle we obtain the minimal surface we want. We can calculate the area of this surface in AdS. (5.J. Note that there is a conformal transformation in the ﬁeld theory that maps a line to a circle. It is. radial coordinate u [168].5) where we have regularized the area by putting a an IR cutoﬀ at z = in AdS. e2 are two orthonormal vectors in four dimensions (which deﬁne the orientation of the circle) and 0 ≤ z ≤ a. (5. the minimum area surface is clearly a plane that intersects the boundary and goes all the way to the horizon (which is just a point on the boundary in the Euclidean case). and we get a contribution to the action S∼ R2 1 A= 2πα 2πα a dθ dza R2 = z2 α a −1 . Subtracting the divergent term we get W ∼ e−S ∼ eR 2 /α =e √ 4πgs N . and let us work in the Poincar´ coore dinates. the Legendre transformation simply subtracts the divergent term r|C|.6) This is independent of a as required by conformal invariance. However. in this case it is easier to use conformal invariance. which is equivalent to a UV cutoﬀ in the ﬁeld theory [66]. Maldacena: Lectures on Large N Field Theories and Gravity 365 Fig. Take C to be a circle of radius a on the boundary. This is because these string coordinates obey the Neumann boundary conditions rather than the Dirichlet conditions. 3. When the loop is smooth. .15) for AdS5 .

Since the action does not depend on x explicitly the solution satisﬁes 1 z2 (∂x z)2 + 1 = constant. Here “quark” means an inﬁnitely massive W-boson connecting the N branes with one brane which is (inﬁnitely) far away. (5. we should consider in this case a D-string worldsheet instead of a fundamental string worldsheet.10) where z0 is determined by the condition L = z0 2 8 All 1 0 dyy 2 1 − y4 = z0 2π 3/2 · Γ(1/4)2 √ (5. This case is related by electric-magnetic duality to the previous case. The classical action for a string worldsheet is S= 1 2πα dτ dσ det(GMN ∂α X M ∂β X N ). which means that the two ends of the string are at the same point in S 5 (one could consider also the more general case with a string ending at diﬀerent points on S 5 [166]). In this case we consider a conﬁguration which is invariant under (Euclidean) time translations. Since we are interested in a static conﬁguration we take τ = t. (5.2) it is possible to compute the quark-antiquark potential in the supergravity approximation [166.11) integrals in this section can be calculated in terms of elliptic or Beta functions. Note that the factors of α cancel out in (5. We take both particles to have the same scalar charge. . and then the action becomes S= T R2 2π L/2 dx −L/2 (∂x z)2 + 1 · z2 (5. Using (5. σ = x. Since we identify the electric-magnetic duality with the SL(2.7) where GMN is the Euclidean AdS5 × S 5 metric. which by symmetry occurs at x = 0. 167]. We put the quark at x = −L/2 and the anti-quark at x = L/2.8) We need to solve the Euler-Lagrange equations for this action. Gauge Theory and Strings We could similarly consider a “magnetic” Wilson loop. We get the same result as in (5. which is also called a ’t Hooft loop [169]. we ﬁnd that the solution is8 1 x = z0 z/z0 dyy 2 1 − y4 .7).6) but with gs → 1/gs . Z) duality of type IIB string theory.366 Unity from Duality: Gravity. (5.9) Deﬁning z0 to be the maximum value of z(x). as they should.

independently of the value of gs . (5. The quark-antiquark energy is given by the diﬀerence of the total length of the strings in a) and b). U= L x U=0 (a) U (b) Fig. Some subleading corrections coming from quantum ﬂuctuations of the worldsheet were calculated in [170–172]. 4. Maldacena: Lectures on Large N Field Theories and Gravity 367 The qualitative form of the solution is shown in Figure 4b. This conﬁguration minimizes the action.13) We see that the energy goes as 1/L.10) in (5. 2 z→0.J. and we ﬁnd E = V (L) = − 2 4π 2 (2gYM N )1/2 · Γ( 1 )4 L 4 (5. Note that the energy is proportional to (gYM N )1/2 . as opposed 2 to gYM N which is the perturbative result.12) Now we compute the total energy of the conﬁguration. This indicates some screening of the charges at strong coupling. b) Conﬁguration after we consider the coupling to the U (N ) gauge theory. as in Fig. Notice that the string quickly approaches x = L/2 for small z (close to the boundary).8). a fact which is determined by conformal 2 invariance. We just plug in the solution (5. 4a). L − x ∼ z3 . The above calculation makes sense for all distances L when gs N is large. 8 . subtract the inﬁnity as explained above (which can be interpreted as the energy of two separated massive quarks. a) Initial conﬁguration corresponding to two massive quarks before we turn on their coupling to the U (N ) gauge theory.

In this case the fundamental string (ending on the quark) will attach to the D-string (ending on the monopole). it is associated with the insertion of two local operators at the two points where the trajectory ends.13) but with gYM → 4π/gYM . In the supergravity approximation the zero-brane follows a geodesic. Geodesics in the hyperbolic plane (Euclidean AdS) are semicircles. Other aspects of the gravity calculation of Wilson loops were discussed in [175–179]. the solution for the fundamental string will be half of the solution we had above.e. but in the limit that gYM is small (but still with gYM N large) we get that the monopole-quark potential is just 1/4 of the quark-quark potential. The area of the minimum surface also contains a logarithmic divergence depending on the angle [168]. and the result will be the same as (5. If we compute the action we get a S=m ds = −2mR adz √ . The simplest example would be a zero-brane (i. 5. the perturbative computation in the gauge theory shows a logarithmic divergence with a coeﬃcient depending on the angle at the cusp.15) e−S ∼ e−2mR log a ∼ 2mR . z a2 − z 2 (5. In fact. One can also calculate the interaction between a magnetic monopole and a quark. the dependence on a is very important. Another interesting case one can study analytically is a surface near a cusp on R4 . The resulting potential is a complicated function of 2 gYM [173]. Naively we might have thought that (as in the previous subsection) the answer had to be independent of a due to conformal invariance. This can be understood from the fact that when g is small the D-string is very rigid and the fundamental string will end almost perpendicularly on the D-string.2 Other branes ending on the boundary We could also consider other branes that are ending at the boundary [180]. We ﬁnd a logarithmic divergence when → 0.368 Unity from Duality: Gravity. a . Therefore. In this case. Therefore. Calculations of Wilson loops in the Higgs phase were done in [174]. Gauge Theory and Strings In a similar fashion we could compute the potential between two magnetic monopoles in terms of a D-string worldsheet. In Euclidean space a zero-brane describes a one dimensional trajectory in anti-de-Sitter space which ends at two points on the boundary. and they will connect to form a (1. 1) string which will go into the horizon. a particle) of mass m. proportional to log( /a). leading to a factor of 1/4 in the potential. since it leads to a result of the form 1 (5. If we subtract the logarithmic divergence we get a residual dependence on a.14) where we took the distance between the two points at the boundary to be L = 2a and regulated the result.

2 we explain how a phase transition studied by Hawking and Page in the context of quantum gravity is mapped into a conﬁnement-deconﬁnement transition in the gauge theory. In general. Again. one gets a logarithmic term. In particular one could consider the “Wilson surfaces” that arise in the six dimensional N = (2. 0) theory. On general grounds one expects that the subleading corrections are given by surfaces that end on more than one loop. ending on a two dimensional surface on the boundary of AdS7 .J. 6 Theories at ﬁnite temperature As discussed in Section 3. and the insights we gain from examining the T > 0 physics will be of a more qualitative nature. One can also compute correlation functions involving more than one Wilson loop.1 Construction The gravity solution describing the gauge theory at ﬁnite temperature can be obtained by starting from the general black three-brane solution and taking the decoupling limit of Section 2 keeping the energy density above . this sort of logarithmic divergence arises when the brane worldvolume is odd dimensional [180].1 how the entropy of near-extremal D3-branes comes out identical to the free ﬁeld theory prediction up to a factor of a power of 4/3. 181]. In that case one has to consider a two-brane. 181]. 0) ﬁeld theory [180. This is precisely the large mR limit of the formula (2. so we reproduce in the supergravity limit the 2-point function described in Section 4. One limiting case is when the surfaces look similar to the zeroth order surfaces but with additional thin tubes connecting them. then in Section 6. These thin tubes are nothing else than massless particles being exchanged between the two string worldsheets [165. the quantities that can be most successfully compared between gauge theory and string theory are those with some protection from supersymmetry and/or conformal invariance – for instance. They are no less interesting for that: we shall see in Section 6. Maldacena: Lectures on Large N Field Theories and Gravity 369 which is precisely the result we expect for the two-point function of an operator of dimension ∆ = mR. 6. with a three dimensional worldvolume. dimensions of chiral primary operators. which is proportional to the rigid string action of the two dimensional surface living on the string in the N = (2. and it implies that the expectation value of the corresponding operator depends on the overall scale. Finite temperature breaks both supersymmetry and conformal invariance.12). To leading order in N this will be just the product of the expectation values of each Wilson loop.

and the ADM mass. but large 2 gYM N at ﬁnite temperature is diﬃcult indeed. one identiﬁes the ﬁeld theory volume with the world-volume of the D3-branes.1). and it is expected to be a reliable 2 guide to the entropy of the gauge theory at large N and large gYM N . and one sets the ﬁeld theory temperature equal to the Hawking temperature in supergravity. but nevertheless the two results for the entropy agreed up to a factor of a power of 4/3.2) The supergravity result is at leading order in ls /R. u4 u0 = πT.1) It will often be useful to Wick rotate by setting tE = it. In the canonical ensemble. Gauge Theory and Strings extremality ﬁnite. The ﬁrst computation which indicated that ﬁnite-temperature U (N ) Yang-Mills theory might be a good description of the microstates of N coincident D3-branes was the calculation of the entropy [182. Two equivalent methods to evaluate FSUGRA are a) to use F = E − T S together with standard expressions for the Bekenstein-Hawking entropy. The resulting metric can be written as du2 + dΩ2 5 hu2 ds2 = R2 u2 (−hdt2 + dx2 + dx2 + dx2 ) + 1 2 3 u4 h=1− 0 . and use the relation between the ﬁnite temperature theory and the Euclidean theory with a compact time direction. equation (6. the entropy of near-extremal D3-branes is just the usual Bekenstein-Hawking result. and b) to consider the gravitational action of the Euclidean solution.370 Unity from Duality: Gravity. (6. The analysis of [182] was limited to a free ﬁeld computation in the ﬁeld theory. . On the supergravity side. where temperature and volume are the independent variables. 3 (6. The result is π2 2 N V T 4. There is no problem on the gauge theory side in working at large N . These corrections do not have an interpretation in the context of CFT because they involve R as an intrinsic scale. the Hawking temperature. 183]. 8 FSUGRA = − FSYM 4 = FSUGRA . and it would acquire corrections suppressed by powers of T R if we had considered the full D3brane metric rather than the near-horizon limit. S = A/4GN .

4) 2 for large gYM N . but it is only because the authors worked in the microcanonical ensemble: rather than identifying the Hawking temperature with the ﬁeld theory temperature. . The weak coupling result is a straightforward although somewhat tedious application of the diagrammatic methods of perturbative ﬁnite-temperature ﬁeld theory. and one restricts attention 9 The result of [182]. The results are 3 2 g N + .J. The gauge theory computation was performed at zero ’t Hooft coupling. 2π 2 YM ζ(3) 3 45 2 + . respectively. based on the non2 renormalization of the two-point function of the stress tensor. (6. Indeed.. diﬀers superﬁcially from (6. 2 The leading corrections to the limiting value of f (gYM N ) at strong and weak coupling were computed in [184] and [187]. f (gYM N ) = + 2 N )3/2 4 32 (gYM 2 f (gYM N ) = 1 − 2 for small gYM N . The constant term is from one loop.3) 2 where f (gYM N ) is a function which smoothly interpolates between a weak coupling limit of 1 and a strong coupling limit of 3/4. φJ ]2 in the N = 4 Yang-Mills action might be expected to freeze out more and more degrees of 2 freedom as the coupling was increased. whereas the supergravity is supposed to be valid at strong ’t Hooft coupling. Maldacena: Lectures on Large N Field Theories and Gravity 371 with a periodicity in the Euclidean time direction (related to the temperature) which eliminates a conical deﬁcit angle at the horizon9 . which would suggest that f (gYM N ) is monotone decreasing. S 1/4 S SYM = (4/3) SUGRA . that f (gYM N ) should remain ﬁnite at strong coupling.2). The 4/3 factor is a long-standing puzzle into which we still have only qualitative insight. it was suggested in [184] that the leading term in the 1/N expansion of F has the form 2 F = −f (gYM N ) π2 2 N V T 4. and unlike in the 1+1-dimensional case where the entropy is essentially ﬁxed by the central charge. An argument has been given [186]... 6 (6.. the ADM mass above extremality was identiﬁed with the ﬁeld theory energy. The relevant one involves a particular contraction of four powers of the Weyl tensor. It is important now to work with the Euclidean solution. there is no nonrenormalization theorem for the coeﬃcient of T 4 in the free energy. It was pointed 2 out early [185] that the quartic potential gYM Tr[φI . The strong coupling result follows from considering the leading α corrections to the supergravity action. and the leading correction is from two loops.

applied at large N and gYM N where loop and stringy corrections are negligible.24]. but exhibits some phase transition at a ﬁnite value of the ’t Hooft coupling. An example of this was studied in [191] long before the AdS/CFT correspondence. the relation I = βF is regarded as the correct one 2 (see [188]). and correspondingly the Bekenstein-Hawking prescription no longer agrees with the free energy computed as βI where I is the Euclidean action. we will restrict our attention to AdS5 . In keeping with the basic prescription for computing Green’s functions. they rely on analyticity properties of the perturbation expansion which do not seem to be proven for ﬁnite temperature ﬁeld theories. It has been conjectured that the interpolating function f (gYM N ) is not smooth. involves extremizing the supergravity action subject to particular asymptotic boundary conditions. it doesn’t matter what the internal manifold is except insofar . The Weyl curvature comes from the noncompact part of the metric. The arguments in [189. where a free energy in ﬁeld theory is equated (in the appropriate limit) with a supergravity action. When there are two or more solutions competing to minimize ISUGRA . We can think of this as the saddle point approximation to the path integral over supergravity ﬁelds. The action including the α corrections no longer has the Einstein-Hilbert form. It is worth noting however that if the AdS5 geometry is part of a string compactiﬁcation. which is no longer AdS5 but rather the AdSSchwarzschild solution which we will discuss in more detail in Section 6. Multiple classical conﬁgurations are possible because of the general feature of boundary value problems in diﬀerential equations: there can be multiple solutions to the classical equations satisfying the same asymptotic boundary conditions.372 Unity from Duality: Gravity. and subsequently resurrected. Gauge Theory and Strings further to the near-horizon limit. and reinterpreted in [24.2 Thermal phase transition 2 The holographic prescription of [23. 6. String amplitudes (when we can calculate them) render on-shell quantities well-deﬁned. and when there is we should sum e−ISUGRA over the classical conﬁgurations to obtain the saddle-point approximation to the gauge theory partition function. Despite the conceptual diﬃculties we can use some simple intuition about path integrals to illustrate an important point about the AdS/CFT correspondence: namely.2. That path integral is ill-deﬁned because of the non-renormalizable nature of supergravity. generalized. The solution which globally minimizes ISUGRA is the one that dominates the path integral. there can be more than one saddle point in the range of integration. 68] as a conﬁnementdeconﬁnement transition in the gauge theory. In particular. We regard this as an unsettled question. Since the qualitative features are independent of the dimension. there can be a phase transition between them. 190] seem as yet incomplete.

Actually. and the boundary is S 3 × S 1 (which is the relevant space for the ﬁeld theory on S 3 with ﬁnite temperature). This requires β= 2πR2 r+ · 2 2r+ + R2 (6. the period is arbitrary.6) with µ = 0 and again with periodic time. both I(X1 ) and I(X2 ) are inﬁnite. corresponding to thermal boundary conditions. corresponding to thermal (anti-periodic) or supersymmetric (periodic) boundary conditions on fermions. in a saddle-point approximation. Another space with the same boundary which is also a local extremum of (6. t ∼ t + β. Tr(−1)F e−βH . there are two possible spin structures on X1 . this space is S 3 × B 2 . Maldacena: Lectures on Large N Field Theories and Gravity 373 as it ﬁxes the cosmological constant. There is an embedding of the Schwarzschild black hole solution into anti-de Sitter space which extremizes the action I=− Explicitly. but for X1 .6) The radial variable r is restricted to r ≥ r+ . The one used in [68. X1 and X2 make separate saddle-point contributions to the usual thermal partition function. is not only metrically distinct from the ﬁrst (being locally conformally ﬂat). In contrast. the elimination of the conical deﬁcit angle at the horizon ﬁxes the period of Euclidean time. In order to make the comparison of I(X1 ) and I(X2 ) meaningful.5) is given by the metric in (6. 3 f 1 16πG5 d5 x √ g R+ 12 R2 · (6. and the more important one is the one with the smaller Euclidean action. This space. but also topologically B 4 ×S 1 rather than S 3 ×B 2 . But. only X1 contributes. X2 is simply connected and hence admits a unique spin structure. We will call this space X2 . in order to eliminate the conical singularity at r = r+ .J. For X2 . the metric is ds2 = f dt2 + 1 2 dr + r2 dΩ2 . where r+ is the largest root of f = 0. we ﬁx the period of Euclidean . or equivalently the radius R of anti-de Sitter space.5) r2 µ f = 1+ 2 − 2· R r (6. Because the S 1 factor is not simply connected. For the purpose of computing the twisted partition function. 184] is to cut oﬀ both X1 and X2 at a deﬁnite coordinate radius r = R0 . so to compute I(X2 )−I(X1 ) a regulation scheme must be adopted. which we will call X1 . The Euclidean time is periodically identiﬁed.7) Topologically. Tre−βH .

and the path C wraps around the circle. in order to compute W (C) we have to evaluate the partition function of strings with a worldsheet D that is bounded by the loop C. X2 is S 3 ×S 1 . 200]. See also [201–204] for other interesting contributions to the ﬁnite temperature literature. 4G5 (2r+ (6. Deconﬁnement at high temperature can be characterized by a spontaneous breaking of the center of the gauge group.8) where again r+ is the largest root of f = 0. and also local thermodynamic instability due to negative speciﬁc heats. In our case the gauge group is SU (N ) and its center is ZN . has the topology B 4 × S 1 . one ﬁnds I(X2 ) − I(X1 ) = 3 2 π 2 r+ (R2 − r+ ) 2 + R2 ) . Since the theory is conformally invariant. Connections with Higgsed states in gauge theory are clearer in [199. The contour C wraps the circle and is not homotopic to zero in X1 .8) (or more precisely its AdS4 analog) can change its sign was interpreted in [191] as indicating a phase transition between a black hole in AdS and a thermal gas of particles in AdS (which is the natural interpretation of the space X1 ). In the deconﬁned phase Fq (T ) is ﬁnite and therefore W (C) = 0. Consider ﬁrst the low temperature phase. Gauge Theory and Strings time on X1 so that the proper circumference of the S1 at r = R0 is the same as the proper length on X2 of an orbit of the Killing vector ∂/∂t. Similar transitions. The expectation value of the Polyakov loop measures the change of the free energy of the system Fq (T ) induced by the presence of the external charge q. An element of the center g ∈ ZN acts on the Polyakov loop by W (C) → g W (C) . This is the expected behavior at low temperatures (compared to the inverse radius of the S 3 ). As discussed in Section 5.374 Unity from Duality: Gravity. which immediately implies that W (C) = 0. In the limit R0 → ∞. W (C) ∼ exp (−Fq (T )/T ). also at r = R0 . The fact that (6. Therefore C is not a boundary of any D. In the gauge theory we interpret this transition as a conﬁnement-deconﬁnement transition. Most of these works are best understood on the CFT side as explorations of exotic thermal phenomena in ﬁnite-temperature gauge theories. as discussed above. the transition temperature must be proportional to the inverse radius of the space S 3 which the ﬁeld theory lives on. The relevance to conﬁnement is explored in [197]. The relevant space is X1 which. The order parameter for the breaking of the center is the expectation value of the Polyakov (temporal) loop W (C) . The black hole is the thermodynamically favored state when the horizon radius r+ exceeds the radius of curvature R of AdS. . The boundary of the spaces X1 . have been studied in the context of spinning branes and charged black holes in [192–198]. In a conﬁning phase Fq (T ) is inﬁnite and therefore W (C) = 0. where the center of the gauge group is not broken.

on R3 one integrates over the local ﬂuctuations of ψ but not over its vacuum expectation value. the solution X2 is not unique and we can add to it an expectation value for the integral of the NS-NS 2-form ﬁeld B on B 2 . Aharony. Phys.M. In the large volume limit the solution corresponds to the N = 4 gauge theory on R3 × S 1 with thermal boundary conditions. Lett. but only in the inﬁnite volume limit (R3 ). J. These notes were made by selected pieces of [1]. and when the S 1 is made small (corresponding to high temperature T ) the theory at distances larger than 1/T eﬀectively reduces to pure Yang-Mills on R3 . Maldacena. B 72 (1974) 461. Indeed. J. Itzhaki. 323 (2000) 183. Nucl. Oz. Phys. Gubser. Some of the non-trivial successes of this approach to QCD are summarized in [1].P. Lett. Park and K. which has the topology S 3 × B 2 . In contrast. Gonzalez-Rey. The contour C is now a boundary of a string worldsheet D = B 2 (times a point in S 3 ). Brandhuber. Das and S. making W (C) = 0 on S 3 . Math. Phys. B 448 (1999) 37. Lett. 2 (1998) 231 [hep-th/9711200]. Nucl. Peeters and K. Lett. Schalm. I. Nucl. I thank O. [11] G. S. B 445 (1998) 142. Phys. Polyakov. B 411 (1997) 59. Lett. Phys. Lett. [5] F. Sonnenschein and S. Ooguri and Y. [4] S.6) was suggested to be holographically dual to a theory related to pure QCD in three dimensions. Ooguri and Y. [6] A. Boonstra. Phys. [7] G. References [1] O. Phys. Phys. Yankielowicz. B 434 (1998) 36. Maldacena. which may be understood as the dependence on the center ZN in the large N limit. [8] A. Adv. ’t Hooft. 205]. Aharony. ’t Hooft. Explicit computations of Polyakov loops at ﬁnite temperature were done in [6. B 79 (1974) 276. Phys. Maldacena: Lectures on Large N Field Theories and Gravity 375 For the high temperature phase the relevant space is X2 . This seems to be in agreement with the fact that in the high temperature phase W (C) = 0 and the center of the gauge group is broken. Oz for collaborating on writing that review. since the center should not be broken in ﬁnite volume (S 3 ). [2] H. H. Sfetsos and K. [3] K.S.J. 75 (1995) 4724 [hep-th/9510017]. Rev. which contributes iψ to the string worldsheet action. [9] J. B 103 (1981) 207. The string theory partition function includes now an integral over all values of ψ. . I also thank the organizers for a very nice school. Polchinski. with vanishing ﬁeld strength. This is an angular parameter ψ with period 2π. Trivedi. Skenderis.J. B. Phys. Theor. Gubser. In [68] the Euclidean black hole solution (6.R. S. This work was supported in part by by the Packard foundation and the DOE grant DE-FGO2-91ER40654. H. Now W (C) = 0 and depends on the value of ψ ∈ U (1). Skenderis. B 517 (1998) 179. Rept. [10] J. It was pointed out in [68] that there is a subtlety with this argument. N. Phys.

71 (1993) 3754 [hep-th/9307049]. J. G. Dhar. Nilsson. Vecchia. E. Gunaydin. Gunaydin and N.P. Math. C. Phys. Dimensional Reduction in Quantum Gravity [gr-qc/9310026]. G.R. J.D. B 272 (1986) 99. Vols. S. Gauge Theory and Strings A. Phys. 1 and 2 (North-Holland. G. Mod. Commun. Phys. Nucl. Gibbons and P. Adv. B 379 (1996) 99 [hep-th/9601029]. Phys. C. B 472 (1996) 591 [hep-th/9602043]. Maldacena and A. Petersen. Klebanov and A. ’t Hooft.M. W. Klebanov. Witten. D 55 (1997) 861 [hep-th/9609026]. Lett. S.N. Phys. private communication (1999). Nucl. 1989). Maldacena. Leigh.S. Phys. I. Rept. Theor. Douglas and S. Phys.R. Gubser and I. Klebanov.K. M. Duﬀ. An Introduction to AdS/CFT correspondence [hep-th/9903007].T. Phys. B 478 (1996) 561 [hep-th/9606185]. Phys. Phys. G. B 360 (1991) 197. Das and S. Phys. Polyakov.H. Nucl.Z. S. Freedman and H.R.R.L. Lett. Gunaydin. Phys. JHEP 12 (1998) 012 [hep-th/9812012]. Rev. B 388 (1996) 51 [hep-th/9605234]. Phys. Phys.D.K.S. Phys. Phys. Phys.R. Strominger. L. B 242 (1984) 377.W. Math. Strominger. Witten. 77 (1996) 2368 [hep-th/9602051].R. Witten. Polyakov.M. E.D.E. D 55 (1997) 5112 [hep-th/9610043]. Missing Math. van Nieuwenhuizen and N. Tseytlin. Nilsson and C. Phys. Gubser. 2 (1998) 253 [hep-th/9802150].K. Supergravities in Diverse Dimensions. Susskind. M. Phys. Seiberg. I. Seiberg. Gubser. Phys. A 4 (1989) 2767. Sezgin. Nucl. B 496 (1997) 231 [hep-th/9702076]. Amsterdam. B 237 (1984) 342. I. D. Horowitz and A. van Nieuwenhuizen and P. B 176 (1986) 45. S. Rev. M. Banks. 20 (1974) 194. Phys. Phys. Vafa. A. Phys. Nucl. Nucl. B 326 (1989) 108. Lett. Randjbar-Daemi. Shenker and L. Pope. Warner. Rev. Rev. Nicolai. Pilch.M.A.G. Phys. 130 (1986) 1. Polchinski. Phys. From three-branes to large N gauge theories [hep-th/9901018]. . Townsend. R. N. G. G. B 428 (1998) 105 [hep-th/9802109]. 36 (1995) 6377 [hep-th/9409089]. M. Warner.E. Klebanov and A. E. G.W. Phys. D 49 (1994) 1912 [gr-qc/9307035]. Callan and J. Lett. Lett. T. Moore. Two lectures on the AdS/CFT correspondence [hep-th/9902022]. K. Rev. P. B 413 (1997) 41 [hep-th/9708005]. Susskind. Rev. Bekenstein.R. Mandal and S. I.G. Witten. Nucl. A. Introduction to the Maldacena conjecture on AdS/CFT [hep-th/9902131]. P. Fronsdal. Elitzur. Klebanov.J. B. Strominger. A. 121 (1989) 351. Lett. Townsend. Horowitz and A. S. Lett. Sierra and P.P. Nucl.T. B 499 (1997) 217 [hep-th/9703040]. B 255 (1985) 63.S. P. Fischler. Lett. E. Wadia. Salam and E. Nucl.376 [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] Unity from Duality: Gravity. J. B 460 (1996) 335-350 [hep-th/9510135]. Mathur. 1998). D 26 (1982) 1988.W. J. Nucl. Nucl. Phys. J.R. S. B. Strominger and C. Schwimmer and N. M. Townsend. JETP Lett. Netherlands. String Theory (Cambridge University Press. A.

Salam. Duﬀ. [63] A. B 430 (1998) 248 [hep-th/9803171].R. B 544 (1999) 218 [hep-th/9810152]. Gutperle. D. Novel supermultiplets of SU (2. E.B. D 58 (1998) 066005 [hep-th/9804174]. Rev. 2|4) and the AdS5 /CFT4 duality [hep-th/9810226]. Nucl.dimensions [hep-th/9604041]. B 224 (1989) 71. [62] S. P. [76] K. Phys. Nucl. [50] E.R. Nucl. [59] M. B 435 (1998) 337 [hep-th/9805131]. The Holographic bound in anti-de Sitter space [hep-th/9805114]. Gonzalez-Rey. B 409 (1997) 239 [hep-th/9705057]. Theor. [61] S. Princeton Series in Physics (Princeton University Press. Phys. B 485 (1997) 85 [hep-th/9608024]. Princeton. Hata and N.J. Douglas and W. [58] M. Nucl. Phys. Lett. [71] O.Y.R. Douglas. Kol. E. [66] L.J. Nucl. Aharony and E. Bergshoeﬀ. B 536 (1998) 149 [hep-th/9804160]. Lett. Phys. Okuyama. the Eleven-Dimensional Supermembrane. Phys. [79] M. H. Lett. Witten. Bagger and J. Pouliot and S. [55] M. Branes in the bulk of Anti-de Sitter space [hep-th/9807225].N. Zagermann.R.S. Tanii. [70] J. Tseytlin. Bergman and B. Green and M. Kabat. [60] M. Pope and E. Gunaydin. B 431 (1994) 484 [hep-th/9408099]. A.H. [65] L. H. Das. B 535 (1998) 83 [hep-th/9804164]. A Note on 1/4 BPS states [hep-th/9902005]. Branes within branes [hep-th/9512077]. [56] M. Phys. D-brane realization of N=2 superYang-Mills theory in four. Das. Susskind and E. Lett. B 377 (1996) 28 [hep-th/9602077]. Taylor. B. [53] C. Shenker. D 59 (1999) 025006 [hep-th/9807164]. A. Seiberg and E. Bergshoeﬀ. B 525 (1998) 104 [hep-th/9712211]. [54] A. Sasakura. Douglas and M. Tong. Nucl. 2 (1998) 505 [hep-th/9803131]. Nucl. Sasakura and S. Sugimoto. Witten. Maldacena: Lectures on Large N Field Theories and Gravity 377 [49] E. Phys. Marcus. Bergshoeﬀ. Li. I. Wess. Lett. Hull and P. 1992). Sezgin and Y. Ferrara. M theory description of 1/4 BPS states in N=4 supersymmetric Yang-Mills theory [hep-th/9811087]. [73] T. [51] E. Witten. Phys. Lett. Kawano and K. [52] M. Phys. [69] F. Holograms of branes in the bulk and acceleration terms in SYM eﬀective action [hep-th/9905037]. Phys. C. Kulik. [57] N. [75] K.K. Witten. JHEP 02 (1999) 012 [hep-th/9901004]. Douglas. Dine and N. Lee and P. Math. B 153 (1985) 392. Sezgin and Y. Hashimoto. Rocek. Sasakura. B 432 (1998) 338 [hep-th/9804139]. [80] M. Yi. Adv. [77] N. . B 205 (1988) 237. [74] O. Andrianopoli and S. Phys. Hashimoto. [78] D. Class. Gunaydin and N. Gubser. Sezgin.M. Rev. B 438 (1995) 109 [hep-th/9410167]. M. Nucl. Seiberg. B 469 (1996) 51 [hep-th/9602064]. Phys. JHEP 11 (1998) 018 [hep-th/9807205]. Kehagias. Minic and M. 2 (1985) L11. Nucl. Salam. Phys. Phys. Supersymmetry and Supergravity. Bergman. Lett. Tanii. [67] O. Lett.A. Grav.R. Hata and N. D. Townsend.R. [72] K. Phys. [81] L. B 431 (1998) 303 [hep-th/9803127]. [68] E. Phys. [64] S. Phys. Phys. B 305 (1988) 497. Park and M. Phys. Quant. Phys. Lett. Phys.J. Romans.

Nucl. Berkooz. Lledo and A. Fronsdal and D. Johansen. Petkova. Flato and C. [84] M. [103] H. Phys. Phys. Osborn and A.J.B. Representations of classical Lie superalgebras. Angelopoulos. [95] S. D 32 (1985) 389. [115] J. The AdS/CFT correspondence and spectrum generating algebras [hep-th/9905097]. Phys. Rev. Phys. Lett.K. [92] M. Minwalla. Aharony. [94] S. Theor. van Nieuwenhuizen. Phys. D 58 (1998) 125025 [hep-th/9808075]. Schalm and R. Zagermann. Phys. [87] C. Rev. Phys. Diﬀerential Geometrical Methods In Mathematical Physics. Lett. Math. Class. Phys. 46 (1998) 157 [hep-th/9806072].G. [85] L. Matusis and L. Marcus. Kim. Rev. Ferrara and C.G. Proc. J. .D. [100] S. C. Phys.R. Ferrara. Correlation functions in the CFTd /AdSd+1 correspondence [hep-th/9804058]. Adv. Nucl. B 97 (1980) 236.A.J. Fronsdal and M. B 433 (1998) 19 [hep-th/9802126]. E. Ferrara and A. Sezgin. Ferrara and C. Lett. D 58 (1998) 105029 [hep-th/9805082]. Grav. Math. Kac. L. Phys. B 483 (1997) 431 [hep-th/9605009]. Deser and A. Grav. Proceedings. [110] G. Flato and C. Nucl. [111] S.B. Dobrev and V. Rev. Math. Romans and P. 597 [97] V. Fortschr. 2/4) Multiplets in the AdS/CFT Correspondence [hep-th/9812067]. Lett. Zaﬀaroni.R. Nucl. Rastelli. Fronsdal. Fronsdal and A. M. Quant. Seiberg.K. B 309 (1993) 279 [hep-th/9302047]. Freedman. JHEP 10 (1998) 004 [hep-th/9808149]. M. 2 (1998) 781 [hep-th/9712074]. Schwimmer. Phys. [106] E. [104] P. Flato. Fronsdal. Phys. Nucl. Nucl. JHEP 07 (1998) 006 [hep-th/9805112]. [102] S. C. 11 (1994) 1387 [hep-th/9308075]. Flato. Grisaru and A. A. 46 (1998) 265 [hep-th/9807150]. Duﬀ. Sternheimer. Phys. Lett. B 491 (1997) 221 [hep-th/9601023]. D. 35 (1987) 537. [112] M. Bulk gauge ﬁelds in AdS supergravity and supersingletons [hep-th/9807090]. M. B 534 (1998) 96 [hep-th/9806042]. Phys. 2 (1978) 421. Minic. [98] V. B 162 (1985) 127. Ferrara. Dobrev and V. Phys. B 143 (1984) 389. Gunaydin and N.S. Witten. Phys. 2 (1985) L19. Osborn. Petkos.J. 22 (1981) 1100. Class. Lett. Suppl. 15 (1998) 2153 [hep-th/9712239]. [96] V. Nicolai and E. Math.Z. Zaﬀaroni.Ii (Berlin 1977). Quant. Klebanov and E. Mathur. Anselmi. 67 (1998) 158 [hep-th/9705117]. [114] D. Class. S. Ferrara. Phys. Ann. [101] S. Seiberg. Chalmers. Petkova. Ferrara. Minic and M. Nucl.378 Unity from Duality: Gravity. Ferrara and C. Fronsdal. Flato and C. [109] D. [86] M. H. B 540 (1999) 247 [hep-th/9805105]. [88] M. Phys. Lett. Grav. Fronsdal. B 172 (1986) 412. [113] H. [93] S. Gimon and D. Erdmenger and H. Lett. Phys. Gubser and I. Zaﬀaroni. D 23 (1981) 1278. Andrianopoli and S. Phys. [99] N. Fronsdal. Phys. Phys. Kutasov and N. Quant. [89] E. Phys. Phys. Klebanov. [107] I. [83] S. Witten. [91] M. Fronsdal. Andrianopoli and S. Berglund. Gunaydin. K. Math. [108] S. 231 (1994) 311 [hep-th/9307010]. Nastase. [105] O. B 536 (1998) 199 [hep-th/9807080]. M. B 532 (1998) 153 [hep-th/9802203]. D. On Short and Long SU (2. [90] H. Lett. Gauge Theory and Strings [82] L. Siebelink.

D. [144] A. Nucl. [126] G. [141] G. [130] W. Kaviani. [119] V.S. Muck and K. D 58 (1998) 106006 [hep-th/9805145]. [124] M. ﬁeld theories in anti-de Sitter space and singletons [hep-th/9803028]. B 533 (1998) 88 [hep-th/9804083]. Theisen and S. Rev. Viswanathan. Tseytlin.Z. Viswanathan. Balasubramanian and P.I.Y. Sfetsos. D 15 (1977) 2752. Chalmers and K. Maldacena: Lectures on Large N Field Theories and Gravity 379 [116] D. Aharony. Counterterms for the Dirichlet prescription of the AdS/CFT correspondence [hep-th/9905046]. Henningson and K.D. B 444 (1998) 92 [hep-th/9810008]. The Graviton in the AdS-CFT correspondence: Solution via the Dirichlet boundary value problem [hep-th/9810151].Z. Lett.N. [117] M. K. Freedman. Kehagias. [121] S. Rev.O. Hawking. Phys. Rivelles. Nucl. Nojiri and S. Lett. Skenderis. D’Hoker and D.S. S. D 59 (1999) 045008 [hep-th/9807098].W. [133] H.S. Nucl. JHEP 07 (1998) 023 [hep-th/9806087].S. Ghezelbash.Z. Yankielowicz. Blau. Lett. [134] E. Viswanathan. Mueck and K. Volovich. Gauge boson exchange in AdSd+1 [hep-th/9809179]. Kraus. Freedman.Z. . J. Phys. Gava. Skenderis. Matusis and L. Seiberg. D’Hoker. Aref’eva and I. Phys. Lee. Henningson and K. Rev. Freedman. Anselmi and A. Rangamani and N. [143] M.W. Frolov. Phys. Chern-Simons theories in the AdS/CFT correspondence [hep-th/9902123]. [140] P. Phys. K. Arutyunov and S. Gibbons and S. [129] H. [138] W. Mathur. Phys. A Note on Anomalies in the AdS/CFT Correspondence [hep-th/9901134]. Rastelli. On subleading contributions to the AdS/CFT trace anomaly [hep-th/9904179].A. Freedman. A. D 59 (1999) 086002 [hep-th/9807097]. Fatollahi. Subleading corrections and central charges in the AdS/CFT correspondence [hep-th/9812092]. [120] W. Johansen. [125] S. Math. Phys. [128] W. Nucl. B 371 (1992) 353. [127] I. Phys. Phys. Muck and K. [142] D. The Large Nc Limit of Four-Point Functions in N = 4 Super Yang-Mills Theory from Anti-de Sitter Supergravity [hep-th/9810051]. On large N conformal theories. B 431 (1998) 63 [hep-th/9803251]. Comments on 4 point functions in the CFT/AdS correspondence [hep-th/9808006]. Phys. Tseytlin. Solodukhin.D.V. [136] E. S. M. Schalm.S. Freedman and W. [135] D. Henningson and K. [139] E. I. Nojiri and S. On the conformal anomaly from higher derivative gravity in AdS/CFT correspondence [hep-th/9903033]. Liu and A. Skiba. Minces and V. Three-Point Green Function of the Stress-Energy Tensor in the AdS/CFT Correspondence [hep-th/9901121]. D. A Stress Tensor for Anti-de Sitter Gravity [hep-th/9902121]. Freedman. Latorre.A. D 58 (1998) 041901 [hep-th/9804035]. Rev. [123] O. 2 (1999) 697 [hep-th/9806074]. Phys. B 526 (1998) 543 [hep-th/9708042].A. Phys. Anselmi. K.T. Holography and the Weyl anomaly [hep-th/9812032]. D’Hoker and D. M. Adv. Viswanathan. Shahrokh AF Tehran Parvizi and A. Minwalla. [131] S.Z.H. Narain and E. Grisaru and A. Liu and A. [118] M. Muck and K. B 435 (1998) 291 [hep-th/9805162]. Johnson and J. [122] D. General scalar exchange in AdSd+1 [hep-th/9811257]. Pawelczyk. B 539 (1999) 403 [hep-th/9806004]. D. Odintsov.Z. S.M. Theor. Rev. Odintsov.J. Phys. [132] S. [137] G.

l’Yi. Lett. [169] G.S. D. Rey and J. Minahan and N.A. 14 (1997) L195 [hep-th/9707044].D. . Solovev. [167] S.A. Graviton and gauge boson propagators in AdS(d+1) [hep-th/9902042]. [154] D. K. Gross and H. [171] S. Gross and H. Quadratic action for Type IIB supergravity on AdS5 × S 5 [hep-th/9811106]. Yee. Ooguri. Forste.E. D. Phys.A. B 441 (1998) 173 [hep-th/9807046]. B 442 (1998) 136 [hep-th/9807223]. Phys. [157] G.S. Naik. Wilson Loops and Minimal Surfaces [hep-th/9904191]. Phys. Phys. Renormalization group ﬂows from holography supersymmetry and a c theorem [hep-th/9904017]. K. Macroscopic strings as heavy quarks in large N gauge theory and anti-de Sitter supergravity [hep-th/9803001]. Phys. Rastelli. Rastelli. Lett. Howe and P. Holographic projection of massive vector ﬁelds in AdS/CFT correspondence [hep-th/9808051].Z. D’Hoker. l’Yi. [151] A.S. Pilch and N. Phys. Stringy corrections to the Wilson loop in N=4 superYang. Phys. Freedman. Freedman. Kogan and O. [165] D. S. D. Rev.S. Maldacena.A. B 544 (1999) 576 [hep-th/9806216]. Lett. Phys. Ooguri. Lechner and D. Holographic normal ordering and multiparticle states in the AdS/CFT correspondence [hep-th/9901144]. D’Hoker. [163] G. Mathur.P. Class. Warner. A. Phys. JHEP 07 (1998) 017 [hep-th/9806140]. [173] J. [147] W. Rev. [175] I. [149] W. Frolov. D. Gauge Theory and Strings [145] G. S. [156] G.Z. Greensite and P.J. Matusis and L.C.A. Schalm. [153] A.E.Z. Polishchuk. [164] E. Lett. Matusis and L. Freedman and L. Arutyunov and S.A. AdS/CFT four point functions: How to succeed at Z integrals without really trying [hep-th/9905049]. [160] E. A. Lechner and M. D’Hoker.380 Unity from Duality: Gravity.-J. Minahan. Nucl.J. Dall’Agata. Schwarz. Nucl. K. l’Yi. [161] E. Ghoshal and S. D 58 (1998) 106002 [hep-th/9805129]. Chalmers and K. Phys. Frolov. Koshelev and O. Sorokin. S. 80 (1998) 4859 [hep-th/9803002]. [174] J. [158] J. B 448 (1999) 218 [hep-th/9811097].H.Z. Adv. Gubser. Drukker. Generating functionals of correlation functions of p form currents in AdS/CFT correspondence [hep-th/9809132]. Dall’Agata. Worldsheet Fluctuations and the Heavy Quark Potential in the AdS/CFT Approach [hep-th/9901057]. Massive symmetric tensor ﬁeld on AdS [hep-th/9905048]. [148] A. [159] P. B 450 (1999) 368 [hep-th/9812238]. West.E. Mathur. Graviton exchange and complete four point functions in the AdS/CFT correspondence [hep-th/9903196]. Liu. D. [146] G. [152] R. [155] G. B 226 (1983) 269. [166] J. 2 (1998) 559 [hep-th/9803111]. Phys.Mills theory [hep-th/9903042]. Frolov.P. Theisen. JHEP 06 (1998) 005 [hep-th/9805104]. ’t Hooft. Freedman. Nucl. JHEP 09 (1998) 022 [hep-th/9809009]. Warner. Rastelli. [170] S. Nucl. [150] W.I. [162] H. Improved heavy quark potential at ﬁnite temperature from Anti-de Sitter supergravity [hep-th/9904147].S. Volovich. Quant.A. Olesen. Lett. Rytchkov. Tonin. Arutyunov and S.S. Theor. Scattering in anti-de Sitter space and operator product expansion [hep-th/9811152]. [172] J. Grav. B 153 (1979) 141. B 238 (1984) 181.C.D. Math. [168] N. Rashkov. Note on the boundary terms in AdS/CFT correspondence for Rarita-Schwinger ﬁeld [hep-th/9904098]. Arutyunov and S.

spinning branes and strongly coupled gauge theories [hep-th/9902195]. Phys. [196] A. W. JHEP 03 (1999) 004 [hep-th/9807196]. On zigzag invariant strings [hep-th/9901131].F. Marolf. [188] R. [185] G.S.V. . Kogan and O. Yee.R. Nucl. Charged AdS Black Holes and Catastrophic Holography [hep-th/9902170]. Gubser. Itzhaki. Gubser. hong Gao and M. Trivedi. [198] M. Klebanov and A. Phys. Gomez and T. B 534 (1998) 202 [hep-th/9805156]. [190] Y. B 541 (1999) 145 [hep-th/9809032]. Phys. Gubser.Commun. Comment on two loop free energy in N=4 supersymmetric Yang. Nucl. Math. A Comment on the entropy of strongly coupled N=4 [hep-th/9904035]. [177] S. Running gauge coupling and quark . Fotopoulos and T. Horowitz and J. Phys. Tseytlin. Birmingham. [179] E. [204] A. [197] M. A 14 (1999) 379 [hep-th/9901143]. R. Taylor. Hunter and M.J. C.M.S. Taylor-Robinson. Page. Li. [182] S. Critical behavior in the rotating D-branes [hep-th/9812121].anti-quark potential from dilatonic gravity [hep-th/9904036]. Wilson loop correlator in the AdS/CFT correspondence [hep-th/9904149]. Ross. Nucl.W. Phys. [205] S. [195] M. [191] S. Polchinski. D 54 (1996) 3915 [hep-th/9602135]. [200] A. Nojiri and S. Witten. Strominger. Berenstein. Wald. Solovev.D. [203] S. Odintsov. Ortin. Hawking. The Operator Product Expansion for Wilson Loops and Surfaces in the Large N Limit [hep-th/9809188]. Tseytlin and S. Theisen and J.T. D 48 (1993) 3427 [gr-qc/9307038]. [187] A.M.M. Phys. F. B 527 (1998) 171 [hep-th/9803135]. Phys. Peet and S.-G.S. Caldarelli and D. Rey. Phys. Phys. Chamblin.W. Rev. I. Lett. [178] K. [201] D. Peet. Soh. Phys. [193] K. S. Landsteiner. B 545 (1999) 217 [hep-th/9806075]. Thermodynamics of spinning D3-branes [hep-th/9810225]. Klemm. [181] D. [184] S.S.C. Alvarez. Yankielowicz.N.P.Mills theory at ﬁnite temperature [hep-th/9811224]. JHEP 03 (1999) 003 [hep-th/9811120].-S.W. Graham and E. Phases of R charged black holes.R. Quant. Nucl. Cvetic and S. [186] N. Cvetic and S. [189] M. 87 (1983) 577. R.A. Mod. [192] S. Cai and K.I. I. Rev.-J. M theory and stringy corrections to Anti-de Sitter black holes and conformal ﬁeld theories [hep-th/9903078]. D 59 (1999) 064005 [hep-th/9811056]. [202] J. Gubser. Rev. C. [183] A. Thermodynamic stability and phases of general spinning branes [hep-th/9903132]. [180] C. Phys. Large N strong/weak coupling phase transition and the correspondence principle [hep-th/9810053]. Maldacena: Lectures on Large N Field Theories and Gravity 381 [176] I. [194] R. Class. Louko and D. D 55 (1997) 6189 [hep-th/9612146]. [199] P. D 59 (1999) 066002 [hep-th/9808081]. Kraus. Corrado. unpublished notes (1997). Rev. 16 (1999) 1197 [hep-th/9808032]. Conformal anomaly of submanifold observables in AdS/CFT correspondence [hep-th/9901021].S.J. Li. Maldacena.W. Zarembo. Emparan. Gubser. Johnson and R.A. Rev. Fischler and J.R.R. Klebanov and A. Hawking and D. Grav. Larsen and S.A. JHEP 12 (1998) 020 [hep-th/9810200]. Myers. C.

Princeton University Princeton.R. .A. KLEBANOV Department of Physics.LECTURE 5 D-BRANES ON THE CONIFOLD AND N = 1 GAUGE/GRAVITY DUALITIES I.S. U. NJ 08544.

. . . . . . . .1 Dimensions of chiral operators . . . 413 6 Infrared physics 6. . . . . .Contents 1 Introduction 385 2 D3-branes on the conifold 388 2. . . . . . . . . . . . . . . . . . . . . 414 414 416 418 . 391 2. . 405 5 Deformation of (KS) the conifold 410 5. . . . . .1 The anomaly as a classical eﬀect in supergravity . 403 4. . .2 SO(4) invariant expressions for the 3-forms . . . . . . . .3 Other ways of wrapping D-branes over cycles of T 1. . . . . . . 412 5. . . . . . . . .2 Wrapped D3-branes as “dibaryons” . . . . . . . . . . .1 Dimensional transmutation and conﬁnement .1 . . . . . . . . . . . . 400 4 The chiral anomaly 402 4. . . . . . . . . . . . . . . . . 393 2.1 The ﬁrst-order equations and their solution . 394 3 The RG cascade 397 3.2 The anomaly as spontaneous symmetry breaking in AdS5 . . . . .1 Matching of the β-functions . . . . . . . . . 6. . . . . . . . 6. . . . .3 Chiral symmetry breaking and gluino condensation . . . . .2 Tensions of the q-strings . . . . . . . . . . . . . . . . .

we describe the gauge/gravity dualities that emerge from placing D3-branes at the apex of the conifold. The associated RR 2-form potential breaks the U (1) R-symmetry to 2M and we study this phenomenon in detail. Springer-Verlag 2002 . Then we show how to break the conformal invariance in this set-up and to introduce logarithmic RG ﬂow into the ﬁeld theory. In particular. we break the conformal symmetry by adding a stack of partially wrapped D5-branes to the system.R. Next. 1 Introduction Comparison of a stack of D3-branes with the geometry it produces leads to a formulation of duality between N = 4 supersymmetric Yang-Mills theory and type II strings on AdS5 × S5 [1–3]. changing the gauge group and introducing a logarithmic renormalization group ﬂow. with discussions of chiral primary operators and wrapped D-branes. which describes the chiral symmetry breaking and conﬁnement in the dual gauge theory. Ouyang Abstract We review extensions of the AdS/CFT correspondence to gauge/ gravity dualities with N = 1 supersymmetry. the eﬀect of these wrapped D5-branes is to turn on the ﬂux of 3-form ﬁeld strengths. These notes are primarily devoted to extensions of the AdS/CFT correspondence to theories with N = 1 supersymmetry. Klebanov. This extra ﬂux also leads to deformation of the cone near the apex. C. We consider ﬁrst the conformal case.D-BRANES ON THE CONIFOLD AND N = 1 GAUGE/GRAVITY DUALITIES I. In the gravity dual. We ﬁrst show how to break some of the supersymmetry without destroying conformal invariance. This may be accomplished through placing a stack of D3-branes at the apex of a Ricci ﬂat 6-dimensional cone [4–7].P. A convenient way to make the coupling constants run logarithmically is to introduce fractional D3-branes c EDP Sciences. It is of obvious interest to consider more general dualities between gauge theories and string theories where some of the supersymmetry and/or conformal invariance are broken. Herzog and P.

the conifold is simple enough that we can follow the program outlined in the paragraph above in great detail. the curved background produced by the stack is ds2 = h−1/2 −dt2 + dx2 + dx2 + dx2 + h1/2 dr2 + r2 dΩ2 .386 Unity from Duality: Gravity. On the other hand. In the gravity dual the eﬀect of these wrapped D5-branes is to turn on the ﬂux of 3-form ﬁeld strengths.13]. such as the chiral anomaly. the duality cascade in the UV. The duality between N = 4 supersymmetric SU (N ) gauge theory and the AdS5 × S5 background of type IIB string theory [1–3] is usually motivated by considering a stack of a large number N of D3-branes. We will start the notes with a very brief review of some of the basic facts about the AdS/CFT correspondence.2) r4 This 10-dimensional metric may be thought of as a “warped product” of the R3. a solution of type IIB supergravity that is dual to a certain N = 1 supersymmetric SU (N + M ) × SU (N ) gauge theory in the limit of strong ’t Hooft coupling. and chiral symmetry breaking and conﬁnement in the IR. The conifold has enough structure that many new aspects of AdS/CFT correspondence emerge that are not immediately visible for the simplest case. The SYM theory is the low-energy limit of the gauge theory on the stack of D3-branes. we review the original AdS/CFT correspondence. the conifold. 1 2 3 5 where dΩ2 5 is the metric of a unit 5-sphere and h(r) = 1 + (1. these fractional branes may be thought of as D5-branes wrapped over 2-cycles in the base of the cone. we consider primarily one particular example of a cone. There are two reasons for this focus. where the conifold is replaced with R6 . This extra ﬂux may lead to deformation of the cone near the apex.1 along the branes and the transverse space R6 . gs (1.4) . For more background the reader may consult. however. the review papers [12. First. the logarithmic running of couplings. and the selfdual 5-form ﬁeld strength is given by F5 = F5 + F5 . (1. Gauge Theory and Strings at the apex of the cone [8–10]. for example. F5 = 16π(α )2 N vol(S5 ). At the same time. which describes the chiral symmetry breaking and conﬁnement in the dual gauge theory [11]. This solution encodes various interesting gauge theory phenomena in a dual geometrical language. is constant.3) The normalization above is dictated by the quantization of Dp-brane tension which implies S8−p Fp+2 = 2κ2 τp N . To make the discussion more concrete. Note that the dilaton.1) L4 · (1. This program eventually leads to the warped deformed conifold [11]. Φ = 0.

11) .: N = 1 Gauge/Gravity Dualities where √ π τp = (4π 2 α )(3−p)/2 κ 2 387 (1. 2 (1. The radial coordinate r is related to the scale in the dual gauge theory.5) and κ = 8π 7/2 gs α is the 10-dimensional gravitational constant.9) L Vol(S ) = 2 κ κ This way we ﬁnd L4 = κN 2 = 4πgs N α 2π 5/2 (1. dr (1. Klebanov et al. (1. 5 z2 (1.3) since the volume of a unit 5-sphere is Vol(S5 ) = π 3 .3) that dr L4 = 4πgs N α .8) A related way to determine the scale factor L is to equate the ADM tension of the supergravity solution with N times the tension of a single D3-brane [14]: √ 2 4 π 5 N. The low-energy limit corresponds to r → 0. for p = 3 we have F5 = (4π 2 α )2 N. (1.7) Then it is not hard to see that the Einstein equation RMN = 2 gs FMPQRS FN PQRS 96 is satisﬁed.10) in agreement with the preceding paragraph. Since −r5 dh = 4L4 . In this limit the metric becomes ds2 = L2 −dt2 + dx2 + dz 2 + L2 dΩ2 . Note that the 5-form ﬁeld strength may also be written as gs F5 = d4 x ∧ dh−1 − r5 dh vol(S5 ).6) S5 which is consistent with (1.I. In particular.R. we ﬁnd by comparing with (1.

Gauge Theory and Strings 2 where z = L . Such theories typically involve product gauge groups SU (N )k coupled to matter in bifundamental representations [16]. S5 . with equal radii of curvature L. In these cases X5 has the local geometry of a 5-sphere. this space is a cone. The simplest example is the Romans compactiﬁcation on X5 = T 1. 4 2 za = 0.17]. Constructions of the dual gauge theories for Einstein manifolds X5 which are not locally equivalent to S5 are also possible. In fact. The dual gauge theory is the IR limit of the world volume theory on a stack of N D3-branes placed at the orbifold singularity of R6 /Γ.1 . and the 5-dimensional sphere. Consider a stack of D3-branes placed at the apex of a Ricci-ﬂat 6-d cone Y6 whose base is a 5-d Einstein manifold X5 . (1. The dual gauge theory is the conformal limit of the world volume theory on a stack of N D3-branes placed at the singularity of a Calabi-Yau manifold known as the conifold [6]. An interesting generalization of the basic AdS/CFT correspondence [1–3] is found by studying branes at conical singularities [4–7]. AdS5 .1 . 18].1) Since this equation is invariant under an overall real rescaling of the coordinates. This describes the direct product of 5-dimensional Anti-de r Sitter space.1 [6.388 Unity from Duality: Gravity. a=1 (2. The simplest examples of X5 are the orbifolds S5 /Γ where Γ is a discrete subgroup of SO(6) [4]. The equality of tensions now requires [15] √ πκN π3 2 4 = 4πgs N α . 2 D3-branes on the conifold The conifold may be described by the following equation in four complex variables. 6 T (2. Comparing the metric with the D-brane description leads one to conjecture that type IIB string theory on AdS5 × X5 is dual to the low-energy limit of the world volume theory on the D3-branes at the singularity.2) . which is a cone over T 1.12) L = 2Vol(X5 ) Vol(X5 ) an important normalization formula which we will use in the following section. the base of this cone is precisely the space T 1. Remarkably. Let us explain this connection in more detail. the metric on the conifold may be cast in the form [18] ds2 = dr2 + r2 ds2 1.1 = (SU (2)× SU (2))/U (1) [6.

7) .1 is 16π [8].3) it is not 3 hard to ﬁnd that the volume of T 1. Using the metric (2. i. Bj .12) it then follows 27 that L4 = 4πgs N (α )2 27κN 27 = · 16 32π 5/2 (2. This quadratic constraint may be “solved” by the substitution zij = Ai Bj .1 .4) whose near-horizon limit is AdS5 × T 1. (2.1 is an S1 bundle over S2 × S2 .I.j 1 zij = √ 2 n σij zn n (2. Here ψ is an angular coordinate which ranges from 0 to 4π. this form of the metric shows that T 1.R.5) The same logic that leads us to the maximally supersymmetric version of the AdS/CFT correspondence now shows that the type IIB string theory on this space should be dual to the infrared limit of the ﬁeld theory on N D3-branes placed at the singularity of the conifold. A simple way to motivate the appearance of the ﬁelds Ai .3) is the metric on T 1.1).6) where σ n are the Pauli matrices for n = 1. Since Calabi-Yau spaces preserve 1/4 of the original supersymmetries we ﬁnd that this should be an N = 1 superconformal ﬁeld theory. while (θ1 . This ﬁeld theory was constructed in [6]: it is SU (N ) × SU (N ) gauge theory coupled to two chiral superﬁelds. N) representation. Ai . From (1. Now placing N D3-branes at the apex of the cone we ﬁnd the metric ds2 = 1+ L4 r4 −1/2 −dt2 + dx2 + dx2 + dx2 1 2 3 1/2 + 1+ L4 r4 dr2 + r2 ds2 1. Bj is to rewrite the deﬁning equation of the conifold. in the (N.1 . The A’s transform as a doublet under one of the global SU (2)s while the B’s transform as a doublet under the other SU (2). φ2 ) parametrize two S2 s in a standard way. N) representation and two chiral superﬁelds. φ1 ) and (θ2 . Klebanov et al. 3 and σ 4 is i times the unit matrix.1 T (2.: N = 1 Gauge/Gravity Dualities where ds2 1. (2. in the (N. 2.1 = T 1 cos θi dφi dψ + 9 i=1 2 2 389 + 1 6 2 2 dθi + sin2 θi dφ2 i i=1 (2. Therefore. as det zij = 0.

In constructing the generalization to the non-abelian theory on N D3-branes. Bl . let us add to the superpotential of this Z2 orbifold a relevant term.8) Therefore. Bj are unconstrained variables. from an N = 1 point ˜ of view. B2 with charges (−1.1 ). l = 1.390 Unity from Duality: Gravity. N). If S5 is described by an equation 6 x2 = 1. 1). then the Z2 acts as −1 on four of the xi and as +1 on the other two. . to ﬁnd the superpotential − g2 [Tr(A1 B1 A2 B2 ) − Tr(B1 A1 B2 A2 )] . The superpotential is ˜ gTrΦ(A1 B1 − A2 B2 ) + gTrΦ(B1 A1 − B2 A2 ). The importance of this choice is that this particular Z2 orbifold of AdS5 × S5 has N = 2 superconformal symmetry. then we ﬁnd a U (1) × U (1) gauge theory coupled to ﬁelds A1 . Now. chiral multiplets Φ and Φ in the adjoint representations of the two U (N )’s. .10) It is straightforward to see what this does to the ﬁeld theory. k. N) representations respectively. From an N = 1 point of view. N) ⊕ (N. If we place a single D3-brane at the singularity of the conifold. m . cancellation of the anomaly in the U (1) R-symmetry requires that the A’s and the B’s each have R-charge 1/2. Using orbifold results for D-branes [16]. Since a marginal superpotential has R-charge equal to 2 it must be quartic. Gauge Theory and Strings where Ai . N) and (N. The model also contains. We simply ˜ integrate out Φ and Φ. x6 . (2. m ˜ TrΦ2 − TrΦ2 . For consistency of the duality it is necessary that we add an exactly marginal superpotential which preserves the SU (2) × SU (2) × U (1)R symmetry of the theory (this superpotential produces a critical line related to the radius of AdS5 × T 1. . . 2 in the (N.9) with real variables x1 . This proposal can be checked in an interesting way by comparing to a certain AdS5 × S5 /Z2 background. i i=1 (2. 2 (2. and the symmetries ﬁx it uniquely up to overall normalization: W = ij kl trAi Bk Aj Bl . this model has been identiﬁed [4] as an AdS dual of a U (N ) × U (N ) theory with hypermultiplets transforming in (N.1 . −1) and B1 . it was proposed in [6] that the SU (N ) × SU (N ) SCFT with this superpotential is dual to type IIB strings on AdS5 × T 1. the hypermultiplets correspond to chiral multiplets Ak . A2 with charges (1.

7].11) [15]. these modes are mixtures of the conformal factors of the AdS5 and S5 and the 4-form ﬁeld. 2. see [19]. Here we discuss the supergravity modes which correspond to chiral primary operators. so the Z2 orbifold with relevant perturbation (2. The exchange of the two U (N )’s is the quantum symmetry of the AdS5 × S5 /Z2 orbifold – the symmetry that acts as −1 on string states in the twisted sector and +1 in the untwisted sector.1 model associated with the conifold. we may keep the discussion of such modes quite general and consider AdS5 × X5 where X5 is any Einstein manifold. Using the fact that the chiral superﬁelds carry R-charge equal to 1/2.1 Dimensions of chiral operators There are a number of further convincing checks of the duality between this ﬁeld theory and type IIB strings on AdS5 × T 1.10) is odd under exchange of the two U (N )’s. (For a more extensive analysis of the spectrum of the model. The twisted sector mode which is a relevant perturbation of the ﬁeld theory is the blowup of the orbifold singularity of S5 /Z2 into the smooth space T 1.: N = 1 Gauge/Gravity Dualities 391 This expression is the same as (2. was given in [7]. which is based on blowing up a Z2 × Z2 orbifold.1 case [15. It is interesting to examine how various quantities change under the RG ﬂow from the S5 /Z2 theory to the T 1. Klebanov et al. Let us try to understand why this works from the point of view of the geometry of S5 /Z2 .10) apparently ﬂows to the T 1. on the ﬁeld theory side it was found that 27 cIR · = cUV 32 (2. the supergravity calculation is in exact agreement with the ﬁeld theory result (2.R. This is a striking and highly sensitive test of the N = 1 dual pair constructed in [6.) For the AdS5 × S5 case. 20].1 .1 ) 32 (2.12) Thus. The behavior of the conformal anomaly (which is equal to the U (1)3 anomaly) was studied R in [15]. The perturbation in (2. Therefore we associate this perturbation with a twisted sector mode of string theory on AdS5 × S5 /Z2 .11) On the other hand.8).1 . . 19. A somewhat diﬀerent derivation of the ﬁeld theory on D3-branes at the conifold singularity. on the supergravity side 27 cIR Vol (S5 /Z2 ) = · = cUV Vol (T 1.I.1 theory. all 3-point functions calculated from supergravity on AdS5 × X5 carry normalization factor inversely proportional to Vol(X5 ). Thus. In fact. The same has been shown to be true for the T 1.

. Aik Bjk ).20].ik Tr(X i1 . and with U (1)R charge k [15. (2. . . k/2). di1 .17) Since the F -term constraints in the gauge theory require that the i and the j indices are separately symmetrized. transforming under SU (2) × SU (2) as (k/2. . . and it seems that the bound (2.14) We will be primarily interested in the modes which correspond to picking the minus branch: they turn out to be the chiral primary ﬁelds.19. .. Here there is a family of wave functions labeled by non-negative integer k. m2 = E + 32 8E 4/5 E (2. X ik ).. in this case we do not encounter operator dimensions lower than 2. 3. Thus.16) In [6] it was argued that the dual chiral operators are tr(Ai1 Bj1 . ∆+ . all chiral primary operators in the N = 4 SU (N ) theory correspond to the conventional branch of dimension. (2. Thus. we ﬁnd that their SU (2)×SU (2)× U (1) quantum numbers agree with those given by the supergravity analysis. It is now well-known that this family of operators with (k) dimensions ∆ = k. For such modes there is a possibility of m2 falling in the range −4 < m2 < −3 (2.15) is satisﬁed for k = 1. First. Gauge Theory and Strings The diagonalization of such modes carried out in [22] for the S5 case is easily generalized to any X5 . this is precisely the special case where the corresponding mode is missing: for k = 1 one of the two mixtures is the singleton [22].1 . Here E = k(k + 4). The absence of k = 1 is related to the gauge group being SU (N ) rather than U (N ). The mixing of the conformal factor and 4-form modes results in the following mass-squared matrix. The situation is diﬀerent for T 1.. is di1 . .ik . The eigenvalues of this matrix are √ m2 = 16 + E ± 8 4 + E. let us recall the S5 case where the spherical harmonics correspond (k) to traceless symmetric tensors of SO(6).392 Unity from Duality: Gravity. However. . The corresponding eigenvalues of the Laplacian are E(k) = 3 k(k + 2) − k2 4 · (2. . k = 2.13) where E ≥ 0 is the eigenvalue of the Laplacian on X5 .15) where there is a two-fold ambiguity in deﬁning the corresponding operator dimension [21].

18) m = V3 κ 9κ .14).2 Wrapped D3-branes as “dibaryons” It is of further interest to consider various branes wrapped over the cycles of T 1. Thus. 20] that the supergravity analysis based on (2.1 and attempt to identify these states in the ﬁeld theory [8]. √ 8π 5/2 L3 π = · (2. and ∆− for the other.I.17). Let us also note that substituting E(1) = 33/4 into (2. While for k > 1 only the dimension ∆+ is admissible. One of these scalar ﬁelds corresponds to the lower component of the superﬁeld Tr(Ai Bj ). 19. For example. For large N the dimensions of such operators calculated from the supergravity are found to be 3N/4 [8]. hence the dimensions of the chiral operators are 3k/2.: N = 1 Gauge/Gravity Dualities 393 In the ﬁeld theory the A’s and the B’s have dimension 3/4.16) we have E(1) = 33/4 which falls within the range (2. while ∆+ = 5/2. the short chiral supermultiplet containing this scalar includes another conformally coupled scalar and a massless fermion [19]. for k = 1 one could pick either branch. wrapped D3-branes turn out to correspond to baryon-like operators AN and B N where the indices of both SU (N ) groups are fully antisymmetrized. in this case we have to pick the unconventional ∆− branch [21]. one encounters an interesting subtlety discussed in [21]. the supersymmetry requires that we pick dimension ∆+ for one of the conformally coupled scalars. from (2.14) we ﬁnd m2 = −15/4 which corresponds to a conformally coupled scalar in AdS5 [22].15).R. which has dimension 3/2. To show how this works in detail. An example of such a 3-cycle is the subspace at a constant value of (θ2 . This is in complete agreement with the fact that the dimension of the chiral superﬁelds at the ﬁxed point is 3/4 and may be regarded as a direct supergravity calculation of an anomalous dimension in the dual gauge theory. In studying the dimensions from the supergravity point of view. the conifold theory provides a simple example of AdS/CFT duality where the ∆− branch has to be chosen for certain operators. therefore. we need to calculate the mass of a D3-brane wrapped over a minimal volume 3-cycle. The mass of the D3-brane wrapped over the 3-cycle is. φ2 ). Since the supersymmetry requires the corresponding dimension to be 3/2. Choosing this branch for k = 1 and ∆+ for k > 1 we indeed ﬁnd following [15. Thus. 2. Here we ﬁnd that ∆− = 3/2. Indeed. In fact. Klebanov et al. while the other corresponds to the upper component which has dimension 5/2.16) reproduces the dimensions 3k/2 of the chiral operators (2. and its volume is found to be V3 = 8π 2 L3 /9 [8]. (2.

we can construct “dibaryon” operators which transform as (1..21) Under the duality these operators map to D3-branes classically localized at a constant (θ1 . 2.. φ1 ). the dimension of the “dibaryon” operators agrees exactly with the supergravity calculation. Z) transform of a D5-brane. Similarly.kN N i=1 Aαiiβi . At the quantum level.βN Dl N i=1 β Bkiiαi .1 (for a complete list.. 9κ 4 (2.. see [23]). B2l = (2.kN β1 . Gauge Theory and Strings For large mL. 2.. which produces a string in AdS5 ..394 Unity from Duality: Gravity. Since the ﬁelds Aα . we will limit explicit statements to D5-branes: since a (p. Thus the SU (2) × SU (2) quantum numbers of B1l are (N + 1. 1). q) 5-brane is an SL(2. a D3-brane may be wrapped over a 2-cycle. Thus.5).1 There are many other admissible ways of wrapping branes over cycles of T 1. For example. The most basic check on the operator identiﬁcation is that. α1 .3 Other ways of wrapping D-branes over cycles of T 1 .. the corresponding operator dimension ∆ approaches mL = 8π 5/2 L4 3 = N. k (2. the existence of two types of “dibaryon” operators is related on the supergravity side to the fact that the base of the U (1) bundle is S2 × S2 ... carry an index α in the N of SU (N )1 and kβ an index β in the N of SU (N )2 . our discussion may be generalized to wrapped (p.βN k Dl 1 . the collective coordinate for the wrapped D3-brane has to be quantized.. and this explains its SU (2)×SU (2) quantum numbers [8]. q) 5-branes using the SL(2. The non-trivial dependence of the tension on the ’t Hooft coupling gs N indicates that such a string is not a BPS saturated object.. In discussing wrapped 5-branes. The tension of such a “fat” string scales as L2 /κ ∼ N (gs N )−1/2 /α .αN k1 .αN β1 . we can construct color-singlet “dibaryon” operators by antisymmetrizing completely with respect to both groups: B1l = α1 . Let us construct the corresponding operators in the dual gauge theory. Z) . k = 1.. This should be contrasted with the tension of a BPS string obtained in [24] by wrapping a D5-brane over RP4 : T ∼ N/α ..19) where in the last step we used (2.. since the exact dimension of the A’s and the B’s is 3/4.20) k where Dl 1 . N + 1).kN is the completely symmetric SU (2) Clebsch-Gordon coeﬃcient corresponding to forming the N + 1 of SU (2) out of N 2’s.

As the string is brought through the membrane. N + 1). a fundamental string stretched between them is created. with its remaining directions ﬁlling R3. Finally. One piece of evidence for this claim is the way the D3-branes wrapped over the S3 behave when crossing the D5-brane domain wall.21). and it is unlikely that this object really exists in the dual CFT. then we ﬁnd a domain wall in the dual ﬁeld theory. ﬁlling out 2 (N. correspondingly.1 then. This is the BPS domain wall which separates adjacent inequivalent vacua distinguished by the phase of the gluino condensate. however. This theory is conﬁning and. the eﬀect in ﬁeld theory is to change the gauge group from SU (N ) × SU (N ) to SU (N + 1) × SU (N + 1). but its tension there is non-vanishing.1 . with its remaining two dimensions parallel to R3.25]. If a D5-brane is wrapped over the entire T 1. the only stable position is at r = 0 which is the horizon. Since each string ends on a charge in the fundamental representation of one of the SU (N )’s. Consider a D5-brane wrapped over the 2-cycle.: N = 1 Gauge/Gravity Dualities 395 symmetry of the type IIB string theory. Consider positioning a “fat” string made of a wrapped D3-brane orthogonally to the domain wall.R. that the domain wall positioned at some arbitrary AdS5 radial coordinate r is not stable: its energy scales as r3 . the resulting ﬁeld theory state is a baryon built out of external quarks. according to the arguments in [24. Therefore. In this case the wrapped D5-brane again falls to the minimum value of the radial coordinate. . Through an analysis of the ﬁve-form ﬂux carried over directly from [24] one can conclude that when one crosses the domain wall. The simplest domain wall is a D3-brane which is not wrapped over the compact manifold. N + 1) ⊕ 2(N. then on the other side the theory is SU (N ) × SU (N + 1) [8]. We should note. In homology there is only one S3 . but for deﬁniteness let us wrap the D3-brane around a particular three-sphere S3 which (1) is invariant under the group SU (2)B under which the ﬁelds Bk transform. it serves as a vertex connecting N fundamental strings. the dual background does not have a horizon. If a D5-brane is wrapped over an S3 . 27].1 . as shown in [26. The origin of this eﬀect is creation of fundamental strings by crossing D5 and D3 branes. We will see. however. The domain wall is tensionless there. The matter ﬁelds Ak and Bk are still bifundamentals. that the domain wall made of a wrapped D5-brane deﬁnitely exists in the SU (N ) × SU (N + M ) generalization of the gauge theory. then it is a domain walls in AdS5 . we show how to construct the SU (N ) × SU (N + M ) theories mentioned above. If this object is located at some ﬁxed r.I. The corresponding state in the SU (N ) × SU (N ) ﬁeld theory is B1 of (2. Klebanov et al. The ﬁeld theory interpretation of a D5-brane wrapped around S2 is more interesting: if on one side of the domain wall we have the original SU (N ) × SU (N ) theory.

In [10] it was conjectured that this solution corresponds to a ﬂow in which the gauge group factors repeatedly drop in size by M units. after a wrapped D3-brane has passed through the wrapped D5-brane domain wall. It was further suggested that the strong dynamics of this gauge theory would resolve the naked singularity in the metric.. until ﬁnally the gauge groups are perhaps SU (2M ) × SU (M ) or simply SU (M ). where the supergravity equations corresponding to this situation were solved to leading order in M/N . and the relative gauge coupling runs logarithmically at all scales.396 Unity from Duality: Gravity. AαN β β or α1 .βN +1 N +1 1 N Aα1 . 9. 9]. but in fact wants to move towards r = 0. . The theory is no longer conformal. decreases logarithmically as well.. The ﬂow is in fact an . as does the superpotential [8. indicating a fundamental of SU (N ). 28. Hence.αN β1 .βN +1 1 N Aα1 . i. .. . one has instead α1 . the D3-brane charge eventually becomes negative and the metric becomes singular.e. the four chiral superﬁelds remain. The addition of M fractional branes at the singular point changes the gauge group to SU (N + M ) × SU (N ). the domain wall is not stable.22) where we have omitted SU (2) indices. Indeed. the conifold suﬀers logarithmic warping. now in the representation (N + M. 27]. These fractional D3-branes are D5-branes wrapped over (collapsed) 2-cycles at the singularity. is free. or the upper index αN +1 . N) and its conjugate. it emerges with a string attached to it due to the string creation by crossing D-branes which together span 8 dimensions [26. This gives a well-deﬁned way of constructing the SUGRA duals of the SU (N ) × SU (N + M ) gauge theories. the singularity is a point. . the logarithm in the solution is not cut oﬀ at small radius.. Instead.. Either the upper index βN +1 . For many singular spaces Y6 there are fractional D3-branes which can exist only within the singularity [8.. The D5-branes wrapped over 2-cycles are examples of a more general phenomenon. 29]. Calculating the tension of a wrapped D5-brane as a function of r shows that it scales as r4 /L2 . In the case of the conifold. indicating a fundamental of SU (N + 1). The D3-brane charge. the relative gauge coupling −2 −2 g1 − g2 runs logarithmically. However. AαN AβN +1 β β α (2. the 5-form ﬂux.. as pointed out in [9]. We will assume that the wrapped D5-branes “fall” behind the horizon and are replaced by their ﬂux in the SUGRA background. Gauge Theory and Strings In the SU (N ) × SU (N + 1) theory. In [10] this solution was completed to all orders.αN β1 . How can this be in supergravity? The answer is simple: the wrapped D3-brane must have a string attached to it..

The warped conifold (KT) solution with such ﬂuxes was constructed in [10].1 = T 1 5 2 1 (g ) + 9 6 4 e2 − e4 √ . in addition to N units of the 5-form ﬂux: 1 4π 2 α S3 F3 = M. e3 ≡ cos ψ sin θ2 dφ2 − sin ψdθ2 . (3. the Einstein metric on T 1. 3 The RG cascade The addition of M fractional 3-branes (wrapped D5-branes) at the singular point changes the gauge group to SU (N + M )× SU (N ).1 .2) (3. 1 (4π 2 α )2 T 1. 2 5 g = e5 . the supergravity dual of this ﬁeld theory involves M units of the 3-form ﬂux. The resulting space. 2 g1 = where e1 ≡ − sin θ1 dφ1 . Once the number of colors in the smaller gauge group is less than M. to Z2 . the resolution of the naked singularity found in [10] occurs through the chiral symmetry breaking of the gauge theory. g2 = (3. leading to conﬁnement [11].1 F5 = N. In short.R. Therefore.I. i=1 (3.4). e5 ≡ dψ + cos θ1 dφ1 + cos θ2 dφ2 . this occurs through the deformation of the conifold. as in pure N = 1 Yang-Mills theory. 2 2 e + e4 g4 = √ . In the supergravity. e2 ≡ dθ1 . It will be useful to employ the following basis of 1-forms on the compact space [30]: e1 − e3 √ . which is broken dynamically. 2 1 e + e3 g3 = √ . a warped deformed conifold.3) (g i )2 .4) . Klebanov et al. e4 ≡ sin ψ sin θ2 dφ2 + cos ψdθ2 . Let us consider the eﬀect on the dual supergravity background of adding M wrapped D5-branes.1 assumes the form ds2 1. In terms of this basis. is completely nonsingular and without a horizon.1) The coeﬃcients above follow from the quantization rule (1. non-perturbative eﬀects become essential.: N = 1 Gauge/Gravity Dualities 397 inﬁnite series of Seiberg duality transformations – a “duality cascade” – in which the number of colors repeatedly drops by M units [11]. The D5-branes serve as sources of the magnetic RR 3-form ﬂux through the S3 of T 1. We will show that these gauge theories have an exact anomaly-free Z2M R-symmetry.

2 ω3 = One can show that [31] (3.1). .10) where 6 is the Hodge dual with respect to the metric ds2 . Note also that gs 6 F3 = H3 . θ1 = θ2 and φ1 = −φ2 .5) H3 = dB2 = where ω2 = 3gs M α dr ∧ ω2 . Thus. Both ω2 and ω3 are closed. (3.398 Unity from Duality: Gravity.11) Note that the self-duality ﬁxes the relative factor of 3 in (3.10) that 2 2 2 gs F3 = H3 .6) 1 1 1 g ∧ g 2 + g 3 ∧ g 4 = (sin θ1 dθ1 ∧ dφ1 − sin θ2 dθ2 ∧ dφ2 ) . F3 = Mα ω3 . 2 B2 = 3gs M α ω2 ln(r/r0 ). Gauge Theory and Strings Keeping track of the normalization factors. (2. Since F3µνλ H3 = 0. 2r (3. 2 (3.7) 2 2 1 5 g ∧ g1 ∧ g2 + g3 ∧ g4 . the quantization condition for RR 3-form ﬂux is obeyed. the RR scalar vanishes as well. As a result. (3. in order to be consistent with the quantization conditions (3. S3 ω3 = 8π 2 (3. gs (3. G3 = F3 − i H3 .2).8) S2 ω2 = 4π.5) (see (2. gs F3 = − 6 H3 .3)).12) µνλ which implies that the dilaton is constant. and the S 3 by θ2 = φ2 = 0. It follows from (3. Φ = 0. (3. the 6 complex 3-form G3 satisﬁes the self-duality condition 6 G3 = iG3 .9) where the S 2 is parametrized by ψ = 0. We will see that this geometrical factor is crucial for reproducing the well-known factor of 3 in the N = 1 beta functions.

24 3 12 3 (3. Thus. Note from (3.1 ). . we may write ˜ F5 = F5 + F5 .R. we ﬁnd that h(r) = 27π(α )2 [gs N + a(gs M )2 ln(r/r0 ) + a(gs M )2 /4] 4r4 (3. This is because ˜ F5 = F5 + B2 ∧ F3 . The non-conservation of the ﬂux is due to the type IIB SUGRA equation ˜ dF5 = H3 ∧ F3 .1 ). F5 = dC4 .19) F5 = 27πα Neﬀ (r)vol(T 1. 5 dr r 6 (3. 2π (3.19) that for a single cascade step Neﬀ (r) → Neﬀ (r) − M the radius changes by a factor r2 /r1 = exp(−2π/3gs M ).13) The solution for the warp factor h may be determined from the trace of the Einstein equation: R= This implies −h−3/2 1 d 5 1 2 (r h ) = H3 .I. agreeing with a result of [32]. Neﬀ (r) → Neﬀ (r) − M which has the eﬀect of decreasing the 5-form ﬂux by M units.14) Integrating this diﬀerential equation.17) and ω2 ∧ ω3 = 54vol(T 1.18) The novel phenomenon in this solution is that the 5-form ﬂux present at the UV scale r = r0 may completely disappear by the time we reach a scale where Neﬀ = 0. and Neﬀ (r) = N + 3 gs M 2 ln(r/r0 ).1 and the conifold: ds2 = h−1/2 (r)dxn dxn + h1/2 (r)(dr2 + r2 ds2 1. 2 (3.20) A related fact is that S2 B2 is no longer a periodic variable in the SUGRA solution once the M fractional branes are introduced: as the B2 ﬂux goes through a period.16) with a = 3/(2π). (3. Klebanov et al.15) 1 1 2 2 2 (H 2 + gs F3 ) = H .1 ).: N = 1 Gauge/Gravity Dualities 399 The 10-d metric found in [10] has the structure of a “warped product” of R3. ˜ An important feature of this background is that F5 acquires a radial dependence [10]. 10 T (3. (3.

1 . this SUGRA solution is reliable for suﬃciently large radii 1. (4π2 α )2 T 1.23) orbifold the- .1 expands slowly toward large r. 2 g1 g2 gs eΦ 1 Exactly the same relations apply to the N = 2 supersymmetric ory [4.1 of the NS-NS and R-R 2-form potentials. Gauge Theory and Strings 1 ˜ Due to the non-vanishing RHS of (3.20). We will work near r = r0 where Neﬀ may be replaced by N .400 Unity from Duality: Gravity. we believe that the continuous logarithmic variation of Neﬀ (r) is related to continuous reduction in the number of degrees of freedom as the theory ﬂows to the IR. T (3. The metric (3. Thus.1 . 2 (3. one notes that the integrals over the S2 of T 1.22) Since T 1. in agreement with (3. the two gauge couplings are determined as follows [6.19). even if gs M is very small. 3. r4 L2 = 9gs M α √ . so that the SUGRA calculation of the β-functions may be compared with SU (Neﬀ + M ) × SU (Neﬀ ) gauge theory. are moduli. 2 2 (3.21) we ﬁnd a purely logarithmic RG cascade: ds2 = r2 L2 ln(r/rs ) dxn dxn + L2 ln(r/rs ) 2 dr + L2 r2 ln(r/rs )ds2 1. the curvatures decrease there so that corrections to the SUGRA become negligible. B2 and C2 . Writing h(r) = L4 ln(r/rs ). We may identify this quantity with Neﬀ deﬁning the gauge group SU (Neﬀ + M ) × SU (Neﬀ ) only at special radii rk = r0 exp(−2πk/3gsM ) where k is an integer. Neﬀ = N − kM . Furthermore. Some support for this claim comes from studying the high-temperature phase of this theory using black holes embedded into an asymptotic KT geometry [33]. 34]. In this regime the separation between the cascade where gs Neﬀ (r) steps is very large. Therefore.1 Matching of the β-functions In order to match the two gauge couplings to the moduli of the type IIB theory on AdS5 ×T 1. The eﬀective number of degrees of freedom computed from the Bekenstein–Hawking entropy grows logarithmically with the temperature.1 F5 is not quantized. 7]1 : 4π 2 4π 2 π + 2 = .13) has a naked singularity at r = rs where h(rs ) = 0. In particular.

13) this gives Λ ∼ r. gs M well-separated. There are diﬀerent ways of establishiing the precise relation.24) S2 1 From the quantization condition on H3 . . These equations are crucial for relating the SUGRA background to the ﬁeld theory β-functions when the theory is generalized to SU (N + M ) × SU (N ) [9.26) Since ln(r/rs ) = ln(Λ/µ).R. A nice review of the derivation of the exact β-functions is in [39]. 2 g1 g2 (3.. 1 so that the cascade jumps are We may consider.I.27) (3. (3. 10]. In gauge/gravity duality the 5-dimensional radial coordinate deﬁnes the RG scale of the dual gauge theory [1–3. 2πα ( S2 B2 ) must be a periodic variable with period 2π. 35. 401 (3. With this convention the additional factor in the β-function does not appear.24) we ﬁnd 8π 2 8π 2 − 2 = 6M ln(r/rs ) + const. Klebanov et al. Let us compare with the Shifman–Vainshtein β-functions [37]2 : d 8π 2 2 d log(Λ/µ) g1 d 8π 2 d log(Λ/µ) g2 2 = 3(N + M ) − 2N (1 − γ). Substituting 2 2 g1 g2 the solution for B2 into (3. (3. 36]. Now we are ready to interpret the solution of [10] in terms of RG ﬂow in the dual SU (N + M ) × SU (N ) gauge theory. (3. this SUGRA result is reliable for any value of gs M provided that gs N 1. As we mentioned earlier.28) 2 These expressions for the β-functions diﬀer from the standard NSVZ form [38] by a factor of 1/(1 − g 2 Nc /8π 2 ). The constancy of the dilaton 2 2 translates into the vanishing of the β-function for 8π + 8π .25) In this section we adopt this UV/IR relation.c.: N = 1 Gauge/Gravity Dualities 4π 2 1 4π 2 gs eΦ = 2 − g2 g1 2πα 2 B2 −π (mod 2π). The diﬀerence comes from the choice of normalization of the vector superﬁelds. This periodicity is crucial for the cascade phenomenon. this choice is dictated by the form of the supergravity action and diﬀers from the canonical normalization by a factor of 1/g 2 . For all metrics of the form (3. for instance.26) implies a logarithmic running of g12 − g12 1 2 in the SU (N + M ) × SU (N ) gauge theory. = 3N − 2(N + M )(1 − γ). which typically corresponds to the Wilsonian renormalization group. The simplest one is to identify the ﬁeld theory energy scale Λ with the energy of a stretched string ending on a probe brane positioned at radius r. We choose the normalization so that the relevant kinetic term in Ê 1 the ﬁeld theory action is 2g2 d4 xd2 θTr(W α Wα )+ h.

because the . the full supergravity solution is only invariant under a Z2M subgroup of this U (1). Although the metric has a continuous U (1)R symmetry. we perform a N = 1 duality transformation on this gauge group factor. namely.402 Unity from Duality: Gravity. the cascade must stop. The SU (N +M ) gauge factor has 2N ﬂavors in the fundamental representation. Thus. there is a scale where the SU (N + M ) coupling. The essential mechanism is the β-functions contribute to the trace anomaly. this becomes an SU (2N − [N + M ]) = SU (N − M ) gauge group. In the previous section we showed how the logarithmic running of the gauge couplings manifests itself in the dual supergravity solution of [10]. however. The conformal invariance of the ﬁeld theory for M = 0. and so we can understand the R-symmetry breaking as an eﬀect of the chiral anomaly. which is related by supersymmetry to the divergence of the U (1)R current. Taking the diﬀerence of the two equations in (3. We may also trace the jumps in the rank of the gauge group to a wellknown phenomenon in the dual N = 1 ﬁeld theory. Anomalies are especially interesting creatures for the gauge/gravity duality. because negative N is physically nonsensical. Gauge Theory and Strings where γ is the anomalous dimension of operators TrAi Bj . As the theory ﬂows to the IR. In the dual quantum ﬁeld theory there are chiral fermions charged under the U (1)R . (3. and symmetry under M → −M . diverges.27). 1 require that γ = − 2 +O[(M/N )2n ] where n is a positive integer [11]. Thus we obtain an SU (N ) × SU (N − M ) theory which resembles closely the theory we started with [11]. This consitutes a geometrical explanation of a ﬁeld theory β-function. To summarize. Under a Seiberg duality transformation. including its normalization. according to (3. Here we show that the chiral anomaly can be read oﬀ the solution as well.29) Remarkably.26) found on the SUGRA side. the coeﬃcient 6M is in exact agreement with the result (3. 4 The chiral anomaly In theories with N = 1 supersymmetry. g1 . To continue past this inﬁnite coupling. and is discussed in Section 5. The necessary modiﬁcation proceeds via the deformation of the conifold. we should not be able to continue the solution (3. the fact that the solution of [10] is singular tells us that it has to be modiﬁed in the IR. The essential observation is that 1/g1 and 1/g2 ﬂow in opposite directions and.27) we then ﬁnd 8π 2 8π 2 − 2 2 g1 g2 = = M ln(Λ/µ)[3 + 2(1 − γ)] 6M ln(Λ/µ)(1 + O[(M/N )2n ]).22) to the region where Neﬀ is negative. Seiberg dual2 2 ity [40]. Tii . ∂i J i . β-functions are related to chiral anomalies [37].

In this section we will study some aspects of the anomaly in detail for the cascading gauge theory.1) This expression is not single-valued as a function of the angular variable β. Indeed. a Z2M subgroup of the U (1) leaves ﬁxed the asymptotic values of the ﬁelds. (4. so that F3 = dC2 is single-valued. the anomaly coeﬃcients computed on each side of the duality agree exactly. This is the R-symmetry of the dual gauge theory. 4. so β → β + is a symmetry precisely if is an integer multiple of π/M . It is crucial.I. and thus corresponds to a symmetry of the system. Under the transformation β → β + . which is a good thing as they do not appear anywhere explicitly in the gravity dual. that the background value of the R-R 2-form C2 does not have this continuous symmetry. with no radiative corrections. although this result is hardly surprising. Finally. F3 is completely independent of β. we may write C2 → M α βω2 . it is a nice check of the duality. even for our non-conformal cascading theory with only N = 1 supersymmetry. C2 → C2 + M α ω2 .1 The anomaly as a classical eﬀect in supergravity The asymptotic UV metric (3. There is no need to appeal to instantons. although F3 is U (1) symmetric. the symmetry breaking is a classical eﬀect on the gravity side. where we can use dual gravity methods to check the calculation. C2 is not U (1)-invariant.: N = 1 Gauge/Gravity Dualities 403 Adler-Bardeen theorem [41] guarantees that anomaly coeﬃcients computed at one loop are exact. but it is single-valued up to a gauge transformation. the R-symmetry is broken spontaneously in the supergravity solution – the bulk vector ﬁeld dual to the R-current of the gauge theory acquires a mass. then extrapolate the results to strong coupling.3) has a U (1) symmetry associated with the rotations of the angular coordinate β = ψ/2.R. Because of the explicit β dependence. a gauge transformation can shift C2 /(4π 2 α ) by an arbitrary integer multiple of ω2 /(4π).13. This Z2M is a symmetry since it respects the asymptotic values of the ﬁelds. 2. there is no smooth global expression for C2 . . Klebanov et al.2) Since S2 C2 is deﬁned modulo 4π 2 α . There are three lessons that we can take away from this analysis [42]. Because is anyway only deﬁned mod 2π. Locally. Second. normalized so that β has period 2π. First. however. The symmetry breaking then appears “anomalous” if one insists on a fourdimensional description. In fact. (4. the signiﬁcance of this fact is that we can compute anomaly coeﬃcients in the ﬁeld theory at weak coupling.

3) S2 where C is the RR scalar. ij 16π 2 ij (4. which vanishes for the case under consideration. A standard result of quantum ﬁeld theory is that in a theory with chiral fermions charged under a global U (1) symmetry of the classical Lagrangian. then under the U (1) rotation (4. derived from supergravity. it is natural that the integral of C2 is dual to the diﬀerence of Θ-angles (it is possible to check this statement explicitly in orbifold backgrounds). ij 16π 2 (4.1 is a modulus.5) are d4 x − ∂i J i + M ˜ ˜ (F a F aij − Gb Gbij ) . (4.6) where J i is the chiral R-current.8) .) As pointed out in [6]. agrees exactly with our expectations from the gauge theory.7) This anomaly equation. Gauge Theory and Strings Let us compare the above analysis with the gauge theory.4) the terms linear in in the dual gauge theory (4. Because the integral of B2 was dual to the diﬀerence of gauge couplings for the two gauge groups.5) If we assume that is a function of the 4 world volume coordinates xi . the integral of the RR 2-form potential C2 over the S2 of T 1. Θ1 + Θ2 ∼ C. we ﬁnd that the small U (1) rotation β → β + induces Θ1 = −Θ2 = 2M . (4. Using the fact that S2 ω2 = 4π. The appearance of the second term is due to the non-invariance of C2 under the U (1) rotation. the Θ terms appear in the gauge theory action as d4 x Θ2 b ˜ bij Θ1 a ˜ aij F F + G G 32π 2 ij 32π 2 ij · (4. The Θ-angles are given by Θ1 − Θ2 = 1 πα C2 .4) With a conventional normalization. the Noether current associated with that symmetry is not generally conserved but instead obeys the equation ∂i J i = 1 32π 2 a ˜ nm Rm Fij F aij m (4. (A similar comparison for the case of an N = 2 orbifold theory appeared in [43].404 Unity from Duality: Gravity. Varying with respect to . we therefore obtain ∂i J i = M a ˜ ˜ Fij F aij − Gb Gbij .

the R-symmetry is global. 48] that the 5-d vector ﬁeld dual to the R-current acquires a mass through the Higgs mechanism.2 The anomaly as spontaneous symmetry breaking in AdS5 Let us look for a deeper understanding of the anomaly from the dual gravity point of view. Rather. Klebanov et al. There R-current conservation was violated not through anomalies but by turning on relevant perturbations or expectation values for ﬁelds. so the anomaly equation as computed from ﬁeld theory is just (4. which must not be anomalous.I. we will ﬁnd that the gauge symmetry is spontaneously broken: the 5-d vector ﬁeld dual to the R-current of the gauge theory “eats” the scalar dual to the diﬀerence of the theta angles and acquires a mass3 . Because the R-symmetry is realized geometrically by invariance under a rigid shift of the angle β.R. Therefore. An equivalent calculation for the SU (N ) gauge group with 2(N + M ) ﬂavors produces the opposite anomaly. as often occurs in the gauge/gravity duality. In the absence of fractional branes there are no background three-form ﬂuxes. The chiral fermions which are their superpartners have Rcharge −1/2 while the gluinos have R-charge 1. We will show that symmetry breaking through anomalies can also have the bulk Higgs mechanism as its dual. Thus. In the case of interest. . or the theory would not make sense at all. the chiral superﬁelds Ai .7). and each one carries R-charge 1/2. No additional fractional D-instanton eﬀects are needed to explain the anomaly. Bj contribute 2N ﬂavors to the gauge group SU (N + M ). Now. it becomes a local symmetry in the full gravity 3 The connection between anomalies in a D-brane ﬁeld theory and spontaneous symmetry breaking in string theory was previously noted in [45] (and probably elsewhere in the literature). The upshot of the calculation presented above is that the chiral anomaly of the SU (N + M ) × SU (N ) gauge theory is encoded in the ultraviolet (large r) behavior of the dual classical supergravity solution. there a are two gauge groups. 4. but in the gravity dual it as usual becomes a gauge symmetry. Similar methods have been used to describe chiral anomalies in other supersymmetric gauge theories [42–44].: N = 1 Gauge/Gravity Dualities 405 where nm is the number of chiral fermions with R-charge Rm circulating in the loop of the relevant triangle diagram. so the U (1) R-symmetry is a true symmetry of the ﬁeld theory. a quantum eﬀect on the gauge theory side turns into a classical eﬀect in supergravity. On the gauge theory side. A closely related mechanism was observed in studies of RG ﬂows from the dual gravity point of view [46. In these cases it was shown [46. so let us deﬁne Fij and Gb to be the ﬁeld strengths ij of SU (N + M ) and SU (N ) respectively. the anomaly M coeﬃcient is 16π2 . 48].

(4. Gauge Theory and Strings theory.10) The equations of motion for the ﬁeld Aµ appear as the χµ components of Einstein’s equations.22]. which is invariant under the combined gauge transformations β → β + λ. An appropriate ansatz to linear order in A is ˜ F5 = dC4 = 1 4 πα N χ ∧ g1 ∧ g2 ∧ g3 ∧ g4 d x ∧ dh−1 + gs 4 −dA ∧ g 5 ∧ dg 5 + 3 L 5 2 dA ∧ dg 5 . It is straightforward to show that ˜ the supergravity ﬁeld equation dF5 = 0 implies that the ﬁeld A satisﬁes the equation of motion for a massless vector ﬁeld in AdS5 space: d 5 dA = 0.406 Unity from Duality: Gravity.9) where h(r) = L4 /r4 . FMPQRS FN 4 · 4! (4.9). and L4 = 27 (4πα gs N ). (4. the unperturbed F5 of (3. 6 r=1 4 (4.14) .11) The ﬁve-form ﬂux will also ﬂuctuate when we activate the Kaluza-Klein ˜ gauge ﬁeld. we can check that the expression for C4 is C4 = 1 −1 4 πα N 1 βg 1 ∧ g 2 ∧ g 3 ∧ g 4 − A ∧ dg 5 ∧ g 5 h d x+ gs 2 2 − 3 −1/4 h 2r 5 2 dA ∧ g 5 . It is convenient to deﬁne 16 the one-form χ = g 5 − 2A. (4. and the associated gauge ﬁelds A = Aµ dxµ appear as ﬂuctuations of the ten-dimensional metric and RR four-form potential [19.12) The ﬁve-dimensional Hodge dual 5 is deﬁned with respect to the AdS5 metric ds52 = h−1/2 dxn dxn + h1/2 dr2 . A → A + dλ. RMN = 2 gs ˜ ˜ PQRS .18) is not self-dual with respect to the gauged metric (4. indeed. (4. The natural metric ansatz is of the familiar Kaluza-Klein form: ds2 = h(r)−1/2 (dxn dxn ) +h(r) 1/2 2 r 1 5 dr2 g − 2A + r2 9 2 2 1 + (g r )2 .13) Using the identity dg 5 ∧ dg 5 = −2g 1 ∧ g 2 ∧ g 3 ∧ g 4 .

16) so that F3 is invariant under the gauge transformation β → β +λ. To do this calculation precisely. θ → θ−λ. a massive vector ﬁeld would satisfy ∂i W i = 0. A more heuristic approach is to consider the type IIB supergravity action to . Let us also deﬁne Wµ = Aµ + ∂µ θ. Another way to see that the vector ﬁeld becomes massive is to compute its equation of motion. To restore the gauge invariance.I.17) we can immediately see how the anomaly will appear in the gravity dual. To a four dimensional observer. the new wrinkle is that the RR three-form ﬂux of (3. 2 (4. Now.: N = 1 Gauge/Gravity Dualities 407 Another way to see that A is a massless vector in AdS5 is to consider the Ricci scalar for the metric (4.10).9) R = R(A = 0) − h1/2 r2 Fµν F µν 9 (4.15) so that on reduction from ten dimensions the ﬁve-dimensional supergravity action will contain the action for a massless vector ﬁeld.R. Thus in the ﬁeld theory one cannot interpret the U (1) symmetry breaking as being spontaneous. and the additional Wr term in (4. Klebanov et al. F3 = Mα (χ + 2W ) ∧ ω2 .18) 5 which is just what one would expect for a massive vector ﬁeld in ﬁve dimensions. In terms of the gauge invariant forms χ and W = Wµ dxµ . the 5-branes introduce M units of RR ﬂux through the threecycle of T 1. Assuming that the NS-NS three form is still given by (3.17) From (4.6). and also ﬁnd the appropriate expressions for the ﬁve-form and metric up to quadratic order in gs M and linear order in ﬂuctuations. As described in Section 2. however. we should derive the χµ components of Einstein’s equations.1 .5) is not gauge-invariant with respect to shifts of β (4. we introduce a new ﬁeld θ ∼ S 2 C2 : F3 = dC2 = Mα 2 g 5 + 2∂µ θdxµ ∧ ω2 (4. This approach is somewhat nontrivial. we ﬁnd that up to terms of order gs M 2 /N the three-form equation implies d W =0 ⇒ L2 1 ∂ W i + 5 ∂r r5 Wr = 0 2 i r r (4. The story changes when we add wrapped D5-branes.18) appears in four dimensions to be an anomaly.

ignoring the 5-form ﬁeld strength contributions: S = ∼ − − − 1 2 2κ10 1 2 2κ10 d10 x d10 x 2 −G10 R10 − −G10 − gs 2 |F3 |2 + . ignores the subtlety of the type IIB action in presence of the self-dual 5-form ﬁeld. Doing this carefully we ﬁnd ˜ Gµν dxµ dxν = hr4 /L4 5/6 (h−1/2 ηij dxi dxj + h1/2 dr2 ).22) 2h3/2 r6 This result. hr4 This is clearly the action for a massive four-dimensional vector ﬁeld. 2 (4.20) gs M α 2 36 Wµ W µ + .20). which has as its equation of motion ∂µ (hr7 F µν ) = m2 hr7 W ν ˜ ˜ which in diﬀerential form notation is d(h7/4 r7 5 dW ) = −m2 h7/4 r7 From the action (4. (4. 81 · (4.2) we must normalize the 5-d metric so that the graviton has a canonical kinetic term.408 Unity from Duality: Gravity.19) h1/2 r2 Fµν F µν 9 (4.23) This shows that the 10-d mass actually appears at a higher order in perturbation theory compared to the result (4. gives instead the following equation for the transverse vector modes: (9M α )4 1 ∂ hr7 ∂r + h∂i ∂i − 7 r hr 64h2 r10 Wi = 0. . . Let us compare this result to earlier work. In [46–48] it was shown that the 5-d vector ﬁeld associated with a U (1)R symmetry acquires a mass in the presence of a symmetry-breaking relevant perturbation. . and that this mass is related in a simple way to the warp factor of the geometry4 . To relate the 5-d metric (4. we see that the mass-squared is given by m2 ˜ = (gs M α ) 2 (4. DeWolfe and K.25) 4 We are grateful to O. It is conventional to write the 5-d gauged supergravity metric in the form ˜ Gµν dxµ dxν = e2T (q) ηij dxi dxj + dq 2 .24) to the 10-d metric (5.22) that ignores the mixing with the 5-form.21) 5 W. which takes the mixing into account. Skenderis for pointing out the relevance of this work to the present calculation. . . however. (4. 12 (4. A more precise calculation [49]. Gauge Theory and Strings quadratic order in W .24) The result of [46] is that m = −2T .

31) Therefore.30) Using the AdS/CFT correspondence (perhaps naively. so now computing the mass-squared by the prescription of [46] we obtain m2 = (gs M )2 4 · α (3π)3/2 (gs N )3/2 (4. The radial variables q and r are related. From (4. at leading order in gs M 2 /N .23) after a rescaling [49] Vi = (hr4 /L4 )2/3 Wi .27) it follows that (mL)2 = 2(gs M )2 · π(gs N ) (4. For these calculations it is convenient to work with the transverse 4-d vector modes Vi and to decouple the longitudinal modes such as Vr . spontaneously breaking the gauge symmetry.1 through the usual Kaluza–Klein reduction. ∂q ∂q (4.25). Klebanov et al. this equation follows from (4. as advertised. as the KT metric is not asymptotically AdS) we ﬁnd that the dimension of the current J µ dual to the vector ﬁeld W µ is ∆=2+ 1 + (mL)2 .26) We can also show that −2T = −2 log(r)+(terms which do not aﬀect the mass to leading order in gs M 2 /N ). the anomalous dimension of the current is ∆ − 3 ≈ (mL)2 /2 = (gs M )2 · π(gs N ) (4.: N = 1 Gauge/Gravity Dualities 5/6 409 The factor hr4 /L4 arises due to the radial dependence of the size of T 1.R. (4.28) In fact.27) where this mass applies to a vector ﬁeld V with a canonical kinetic term for the metric (4.I.32) . in contrast to some earlier examples [3. It is interesting that the anomaly appears as a bulk eﬀect in AdS space. by log(r) ∼ q gs M 2 − L 2πN q L 2 · (4. 50] where anomalies arose from boundary terms.29) The nonvanishing vector mass is consistent with gauge invariance because the massless vector ﬁeld A has eaten the scalar ﬁeld θ. The equation of motion of V is e−2T ∂ 2T ∂ e + e−2T ∂i ∂i − m2 Vi = 0. (4. The appearance of a mass implies that the R-current operator should acquire an anomalous dimension.

2) where ds2 is the metric of the deformed conifold (5. but with the conifold replaced by the deformed conifold as the transverse space.3) 2 .1) should be replaced by the deformed conifold 4 2 zi = ε2 . 5 Deformation of (KS) the conifold It was shown in [11] that. (5. to remove the naked singularity found in [10] the conifold (2. Our supergravity calculation predicts that this anomalous dimension is corrected at large gs N by an extra factor of 1/(gs N ). 10 6 (5. and it must be invariant under the map M → −M . This is the same 6 type of “D-brane” ansatz as (3. Gauge Theory and Strings We can obtain a rough understanding of this result by considering the relevant weak coupling calculation in the gauge theory.1) in which the singularity of the conifold is removed through the blowing-up of the S3 of T 1. i=1 (5. The leading correction to the current-current two-point function comes from the three-loop Feynman diagram composed of two triangle diagrams glued together. N → N + M .3). the lowest order piece of the anomalous dimension will be of order (gs M )2 . The 10-d metric of [11] takes the following form: ds2 = h−1/2 (τ )dxn dxn + h1/2 (τ )ds2 . Of course. 30. The metric of the deformed conifold was discussed in some detail in [18. It is diagonal in the basis (3.410 Unity from Duality: Gravity.1 . and the resulting anomalous dimension γJ is quadratic in M and N .13).2): ds2 = 6 τ 1 1 4/3 ε K(τ ) (dτ 2 + (g 5 )2 ) + cosh2 [(g 3 )2 + (g 4 )2 ] 2 3K 3 (τ ) 2 + sinh2 where K(τ ) = (sinh(2τ ) − 2τ )1/3 · 21/3 sinh τ (5. which simply interchanges the two gauge groups. γJ must vanish when M = 0.4) τ [(g 1 )2 + (g 2 )2 ] . Thus. it would be interesting to understand this result better from the gauge theory point of view. 51].

The additional two directions. shrink as 1 4/3 ε (2/3)1/3 τ 2 [(g 1 )2 + (g 2 )2 ].8) (5. 2 (5.R. 30]. and B2 = gs M α [f (τ )g 1 ∧ g 2 + k(τ )g 3 ∧ g 4 ]. We have F5 = B2 ∧ F3 = where = f (1 − F ) + kF.9) (5. 2 2 (5. 8 The simplest ansatz for the 2-form ﬁelds is F3 = = Mα 2 Mα 2 g 5 ∧ g 3 ∧ g 4 + d[F (τ )(g 1 ∧ g 3 + g 2 ∧ g 4 )] g 5 ∧ g 3 ∧ g 4 (1 − F ) + g 5 ∧ g 1 ∧ g 2 F +F dτ ∧ (g 1 ∧ g 3 + g 2 ∧ g 4 ) . and F5 = 4gs M 2 (α )2 ε−8/3 dx0 ∧ dx1 ∧ dx2 ∧ dx3 ∧ dτ K 2 h2 (τ ) · sinh2 (τ ) (5. Klebanov et al. with F (0) = 0 and F (∞) = 1/2.5) and in terms of this radial coordinate ds2 → dr2 + r2 ds2 1. 2 = gs M α dτ ∧ (f g 1 ∧ g 2 + k g 3 ∧ g 4 ) 2 1 + (k − f )g 5 ∧ (g 1 ∧ g 3 + g 2 ∧ g 4 ) . 411 (5.13) (5. 6 T At τ = 0 the angular metric degenerates into dΩ2 = 3 1 4/3 1 ε (2/3)1/3 (g 5 )2 + (g 3 )2 + (g 4 )2 .10) ˜ As before. the self-dual 5-form ﬁeld strength may be decomposed as F5 = F5 + F5 .: N = 1 Gauge/Gravity Dualities For large τ we may introduce another radial coordinate r via r2 = 3 25/3 ε4/3 e2τ /3 .7) H3 = dB2 (5.6) which is the metric of a round S3 [18.12) gs M 2 (α )2 (τ )g 1 ∧ g 2 ∧ g 3 ∧ g 4 ∧ g 5 . corresponding to the S2 ﬁbered over the S3 .11) .1 .I. 4 (5.

14). these equations imply the self-duality of the complex 3-form with respect to the metric of the deformed conifold: 6 G3 = iG3 . the warp factor may be determined by integrating (5.17) Now that we have solved for the 3-forms on the deformed conifold.15) = (1 − F ) tanh2 (τ /2). 4 sinh2 τ (5. the boundary condition that h vanishes.1 The ﬁrst-order equations and their solution In searching for BPS saturated supergravity backgrounds. In fact. Gauge Theory and Strings 5. = 2 (5.18) This behaves as τ 3 for small τ . The solution is F (τ ) f (τ ) k(τ ) = = = sinh τ − τ .14) These equations follow from a superpotential for the eﬀective radial problem [52]. = F coth2 (τ /2). 1 (k − f ). K 2 (τ ) sinh2 τ (5.19) .16) f (1 − F ) + kF . For large τ we impose.412 Unity from Duality: Gravity. 2 sinh τ (5.15). 2 sinh τ τ coth τ − 1 (cosh τ + 1). this is possible for our ansatz [11]: f k F and h = −α where α = 4(gs M α )2 ε−8/3 . form a closed system and need to be solved ﬁrst. as usual. 2 sinh τ τ coth τ − 1 (cosh τ − 1). The resulting integral expression for h is h(τ ) = α 22/3 I(τ ) = (gs M α )2 22/3 ε−8/3 I(τ ). (5. 4 (5. the second order equations should be replaced by a system of ﬁrst-order ones. Note that the ﬁrst three of these equations (5. Luckily. First we note that (τ ) = f (1 − F ) + kF = τ coth τ − 1 (sinh 2τ − 2τ ).

.: N = 1 Gauge/Gravity Dualities where I(τ ) ≡ ∞ 413 dx τ x coth x − 1 (sinh(2x) − 2x)1/3 . for large gs M the curvatures found in our solution are small everywhere. Thus.21). The small τ behavior follows from the convergence of the integral (5. matches the form of the large-τ solution (3. and lifting the result to M-theory maps our background to a Becker-Becker solution supported by a G4 which is a (2.R. Klebanov et al. 1) form on the CY space.I.23) 1 +(g 3 )2 + (g 4 )2 + τ 2 [(g 1 )2 + (g 2 )2 ] · 4 This metric will be useful in Section 6 where we investigate various infrared phenomenon of the gauge theory. the curvatures are small and the SUGRA approximation is reliable. 2) form on T 2 × CY. 1 4 (5. 54] it was shown that the warped background of the previous section preserves N = 1 SUSY if and only if G3 is a (2.5). Perhaps the easiest way to see the supersymmetry of the deformed conifold solution is through a T -duality.19).22) where a0 ≈ 0. Performing a T -duality along one of the longitudinal directions.2 SO(4 ) invariant expressions for the 3-forms In [53. but it is not hard to see that I(τ → 0) → a0 + O(τ 2 ).1 times the deformed conifold: ds2 10 → ε4/3 1/2 21/3 a0 gs M α dxn dxn + a0 6−1/3 (gs M α ) 1/2 1 2 1 5 2 dτ + (g ) 2 2 (5.21) I(τ → ∞) → 3 · 2−1/3 τ − e−4τ /3 . This is true even far in the IR. G-ﬂux of this type indeed produces a supersymmetric background [55]. for small τ the ten-dimensional geometry is approximately R3.20) We have not succeeded in evaluating this integral in terms of elementary or well-known special functions. 5. This is the ’t Hooft coupling of the gauge theory found far in the IR. This I(τ ) is nonsingular at the tip of the deformed conifold and. As long as this is large. from (5. while at large τ the integrand becomes ∼ xe−4x/3 . (5. sinh2 x (5. Very importantly. since the radius-squared of the S3 at τ = 0 is of order gs M in string units.71805.

While the singular conifold has no dimensionful parameter. Gauge Theory and Strings While writing G3 in terms of the angular 1-forms g i is convenient for some purposes. ¯ z (5. (6.1 Dimensional transmutation and conﬁnement The resolution of the naked singularity via the deformation of the conifold is a supergravity realization of the dimensional transmutation. 6 Infrared physics We have now seen that the deformation of the conifold allows the solution to be non-singular. 6. i = 1. That G3 is indeed (2.24) ¯ z We also note that the NS-NS 2-form potential is an SO(4) invariant (1. 1) nature of the form is not manifest. we saw that turning on the R-R 3-form ﬂux produces the logarithmic warping of the KT solution. that there is a gluino condensate that breaks the Z2M chiral symmetry down to Z2 and that there are domain walls separating inequivalent vacua.414 Unity from Duality: Gravity.25) The derivation of these formulae is given in [31]. In particular. Other stringy approaches to infrared phenomena in N = 1 SYM theory have recently appeared in [57–59].1) . 1) form: B2 = igs M α τ coth τ − 1 2ε4 sinh2 τ ijkl zi zj dzk ∧ d¯l . we will now demonstrate that there is conﬁnement. Below we write the G3 found in [11] in terms of the obvious 1-forms on the deformed conifold: dz i and d¯i . 1) was demonstrated in [56] with the help of a holomorphic basis. 3. and so is the metric. In the following sections we point out some interesting features of the SUGRA background we have found and show how they realize the expected phenomena in the dual ﬁeld theory. From (5. 4: z G3 = Mα 2ε6 sinh4 τ sinh(2τ ) − 2τ ( sinh τ ijkl zi zj dzk ∧ d¯l ) ∧ (¯m dzm ) ¯ z z +2(1 − τ coth τ )( ijkl zi zj dzk ∧ dzl ) ∧ (zm d¯m ) · (5. that the theory has glueballs and baryons whose mass scale emerges through a dimensional transmutation. the (2. 2. Our expressions for the gauge ﬁelds are manifestly SO(4) invariant. The scale necessary to deﬁne the logarithm transmutes into the the parameter ε that determines the deformation of the conifold.5) we see that ε2/3 has dimensions of length and that τ = 3 ln(r/ε2/3 ) + const.

I.21) should be identiﬁed with ε2/3 . A wrapped D3-brane plays the role of a baryon vertex which ties together M fundamental strings. 61]. On the other hand. If the contour has a very large area A. Consider a Wilson contour positioned at ﬁxed τ . The minimal area surface bounded by the contour bends towards smaller τ . From the fact that the coeﬃcient of dxn dxn is ﬁnite at τ = 0. the connection between these two objects becomes clearer when one notes that for M > 0 the dibaryon has M uncontracted indices. We may interpret various branes wrapped over this 3-cycle in terms of the gauge theory. This should be contrasted with the AdS5 metric where this function vanishes at the horizon. A simple estimate shows that the string tension scales as Ts = 1 1/2 24/3 a0 π ε4/3 · (α )2 gs M (6. then most of the minimal surface will drift down into the region near τ = 0. (6. Studying a probe D3-brane in the background of our solution show that the mass of the baryon scales as Mb ∼ M ε2/3 · α (6. and calculate the expectation value of the Wilson loop using the prescription [60. Note that for M = 0 the D3-brane wrapped on the S3 gave a dibaryon [8]. and therefore joins M external charges.23) makes it clear that the dynamically generated 4-d mass scale. the scale rs entering the UV solution (3. and so the resulting Wilson loop satisﬁes the area law. we see that Ts ∼ gs M (mglueball )2 .: N = 1 Gauge/Gravity Dualities 415 Thus. the form of the IR metric (5.5) Due to the deformation. Klebanov et al. Note that the 3-cycle has the minimal volume near τ = 0.4) Comparing with the string tension. hence all the wrapped branes will be localized there.6) . the full SUGRA background has a ﬁnite 3-cycle. or with the singular metric of [10] where it blows up.3) We will return to these conﬁning strings in the next section. which sets the tension of the conﬁning ﬂux tubes. is ε2/3 √ · α gs M (6. we ﬁnd that a fundamental string with this surface will have a ﬁnite tension.2) The reason the theory is conﬁning is that in the metric for small τ (5.23) the function multiplying dxn dxn approaches a constant. The masses of glueball and Kaluza-Klein (KK) states scale as mglueball ∼ mKK ∼ ε2/3 · gs M α (6.R.

Tq+q < Tq + Tq . It is interesting to ask how the tension of this class of conﬁning strings depends on q. 2. An interesting generalization is to consider Wilson loops in antisymmetric tensor representations with q indices where q ranges from 1 to M − 1. To explain what they are. M q = 1.1 . for brevity we build on a closely related result of Bachas et al. (6. Here the conﬁning q-string is described by q coincident fundamental strings placed at τ = 0 and oriented along the R3. M − 1 (6.8) then the q-string will not decay into strings with smaller q. This is precisely the situation found by Douglas and Shenker (DS) [62] in softly broken N = 2 gauge theory. .7) in the limit of large area. This type of behaviour is also found in the supergravity duals of N = 1 gauge theories [66]. and there is a symmetry under q → M − q which corresponds to replacing quarks by anti-quarks. this R-R ﬂux blows up the q fundamental strings into a D3-brane wrapping an S2 inside the S3 . 65]: Tq = Λ2 sin πq . These Wilson loops can be thought of as conﬁning strings which connect q probe quarks on one end to q corresponding probe anti-quarks on the other. . the appearance of composite conﬁning strings. For q = M the probe quarks combine into a colorless state (a baryon). . hence the corresponding Wilson loop does not have an area law. but it is important that there are M units of F3 ﬂux through the S3 . q = 1 corresponds to the fundamental representation as denoted above. Gauge Theory and Strings 6. let us recall that the basic string corresponds to the Wilson loop in the fundamental representation.416 Unity from Duality: Gravity.5 In the deformed conifold solution analyzed above both F5 and B2 vanish at τ = 0. The classic criterion for conﬁnement is that this Wilson loop obey the area law − ln W1 (C) = T1 A(C) (6. (HSZ) [63] in the MQCD approach to conﬁning N = 1 supersymmetric gauge theory [64.9) where Λ is the overall IR scale. . [68]. and later by Hanany et al. Although the blow-up can be shown directly. .2 Tensions of the q-strings The existence of the blown up 3-cycle with M units of RR 3-form ﬂux through it is responsible for another interesting infrared phenomenon. In the S-dual of 5 Qualitatively similar conﬁning ﬂux-tubes were examined in [67] where the authors use the near horizon geometry of non-extremal D3-branes to model conﬁnement. In fact. If it is a convex function.

11) while the world volume ﬁeld is q F = − sin θdθ ∧ dφ. we ﬁnd ψ → π − ψ. at τ = 0 the metric is ε4/3 2 21/3 a0 gs M α 1/2 1/2 dxn dxn + bM α dψ 2 + sin2 ψdΩ2 . . (6.93266. We are now using the standard round metric on S3 so that ψ is the azimuthal angle ranging from 0 to π. M 2 (6.14) The tension of the wrapped brane is given in terms of the solution of this equation by 4/3 Tq = 2 121/3 π 2 gs α 2 sin ψ 1 + (b2 − 1) cos2 ψ.12) Following [68] closely we ﬁnd that the tension of a D3-brane which wraps an S2 located at the azimuthal angle ψ is sin(2ψ) πq − b sin ψ + ψ − 1/3 π 2 g 2 α 2 b 2 M 12 s 2 4 4/3 2 1/2 . After applying S-duality to the KS solution.1 .10) where b = 2a0 6−1/3 ≈ 0.13) Minimizing with respect to ψ we ﬁnd ψ− πq 1 − b2 = sin(2ψ).15) Note that under q → M − q.: N = 1 Gauge/Gravity Dualities 417 our type IIB gravity model. This is a crucial property needed for the connection with the q-strings of the gauge theory.I.1 × S3 geometry with M units of NS-NS H3 ﬂux through the S3 and q coincident D1-branes along the R3. (6. This geometry is very closely related to the setup of [68] whose authors showed that the q D0-branes blow up into an S2 . The NS-NS 2-form ﬁeld at τ = 0 is B2 = M α ψ− sin(2ψ) 2 sin θdθ ∧ dφ. 2 (6. so that TM−q = Tq . but our probe brane calculation is somewhat diﬀerent from [68] because the radius of our S3 is diﬀerent. at τ = 0 we ﬁnd the R3. 2 (6. (6. Klebanov et al. T -dualizing along the D1-brane direction we ﬁnd q D0-branes on an S3 with M units of NS-NS ﬂux.R. We will ﬁnd the same phenomenon.

As we have shown earlier. all the tensions are multiplied by gs . . Gauge Theory and Strings Although (6. 6. . M k = 1. the tension in the KS case is approximately 96.7% of that in the b = 1 case. .418 Unity from Duality: Gravity. In the large M limit. In Section 3 we showed that the corresponding symmetry of the UV (large τ ) limit of the metric is ψ→ψ+ 2πk . hence after the S-duality we ﬁnd only H3 = dB2 . Our argument applied to the MN background leads very simply to the DS–HSZ formula for the ratios of q-string tensions (6. we note that (1−b2 )/2 ≈ 0.17). this formula also holds approximately for the KS background. then ψ ≈ πq/M and Tq ∼ sin πq · M (6.3 Chiral symmetry breaking and gluino condensation Our SU (N + M ) × SU (N ) ﬁeld theory has an anomaly-free Z2M R-symmetry. M.16) with the naive string tension (6.06507 is small numerically.3). (6.14) is not exactly solvable. The value of B2 at the minimal radius is again given by (6. we expect interactions among the strings to become negligible and the q-string tension to become just q times the ordinary string tension (6. In this background only the F3 ﬂux is present.17). The extra gs appears because we have been computing tensions in the dual background. Tq sin πq M (6. Indeed.11). now we ﬁnd sin πq Tq M = . . When we S-dualize back to the original background with RR-ﬂux and q F -strings. This simpliﬁes the probe calculation and makes it identical to that of [68].16) The deviations from this formulae are small: even when ψ = π/4 and correspondingly q ≈ M/4. 2. It is interesting to note that recent lattice simulations in non-supersymmetric pure glue gauge theory [69] appear to yield good agreement with (6.18) . we ﬁnd that gs Tq = qTs in the large M limit. It is interesting to compare (6. In particular. If we ignore the RHS of this equation.3) we obtained in the previous section. There is a subtle diﬀerence however from the calculation for the KS background in that now the parameter b entering the radius of the S3 is equal to 1. An analogous calculation for the MN background [57] proceeds almost identically.17) without making any approximations.

corresponding to breaking of the Z2M UV symmetry through the IR eﬀects. has the Z2 symmetry generated by ψ → ψ + 2π. a fundamental string can end on the wrapped D5-brane. As a result. The domain walls constructed out of the wrapped D5-branes separate these inequivalent vacua. there are M inequivalent vacua: there are exactly M diﬀerent discrete orientations of the solution. which depends on ψ through cos ψ and sin ψ. The B-cycle is bounded by a 2-sphere at τ = ∞.) They looked for the deviation of the complex 2-form ﬁeld C2 − gis B2 from its asymptotic large τ form that enters the KT solution: δ C2 − i B2 gs ∼ M α −τ τ e [g1 ∧ g3 + g2 ∧ g4 − i(g1 ∧ g2 − g3 ∧ g4 )] 4 . the wrapped D5-brane produces a discontinuity in B F3 . Klebanov et al. If to the left of the domain wall B F3 = 0. It is expected that ﬂux tubes can end on these domain walls [70]. Such a domain wall is obviously a stable object in the KS background and crossing it takes us from one ground state of the theory to another.R. Indeed.21) In supersymmetric gluodynamics the breaking of chiral symmetry is associated with the gluino condensate λλ . Also.18) on the original ﬁeld conﬁguration (4. where B is the cycle dual to the S3 . baryons can dissolve in them. Let us consider domain walls made of k D5-branes wrapped over the ﬁnite-sized S3 at τ = 0.: N = 1 Gauge/Gravity Dualities 419 Recalling that ψ ranges from 0 to 4π. then to the right of the domain wall F3 = 4π 2 α k.1 . as in the basic solution derived in the preceding sections. Indeed.20) for large τ .9) it is clear that to the right of the wall ∆C2 → πα kω2 (6.1). hence B F3 = S2 ∆C2 .I. we see that the full solution. By studying a probe D5-brane in the metric. with remaining directions parallel to R3. we ﬁnd that the domain wall tension is Twall ∼ 1 ε2 · gs (α )3 (6. Therefore from (3. A holographic calculation of the condensate was carried out by Loewy and Sonnenschein in [71] (see also [72] for previous work on gluino condensation in conifold theories. (6.19) B as follows from the quantization of the D5-brane charge. This change in C2 is produced by the Z2M transformation (6.

Lett. The presence of an extra ln(r/ε2/3 ) factor is presumably due to the fact that the asymptotic KT metric diﬀers from AdS5 by such logarithmic factors. Phys. JHEP 0008 (2000) 052 [hep-th/0007191]. Lawrence. [13] I. This work was supported in part by the NSF grant PHY-9802484. Adv. is grateful to S. Nekrasov and C. A. From the angular dependence of the perturbation we see that the dual operator is SU (2) × SU (2) invariant and carries R-charge 1. [9] I. Aharony. Phys. [5] A.M.22) In a space-time that approaches AdS5 a perturbation that scales as r−3 corresponds to the expectation value of a dimension 3 operator. Peet. Tseytlin. 2 (1998) 253 [hep-th/9802150]. Phys. Phys. 2 (1998) 231 [hep-th/9711200]. Phys. Gauge Theory and Strings M α ε2 ln(r/ε2/3 )eiψ (dθ1 − i sin θ1 dφ1 ) ∧ (dθ2 − i sin θ2 dφ2 ). Oz. Gubser. [7] Non-Spherical Horizons.K.R. Kachru and E. Math. B 533 (1998) 199 [hep-th/9803015]. Klebanov and A.420 ∼ Unity from Duality: Gravity. I. References [1] J. 6 It would be nice to understand the relative factor of gs M between Twall and λλ . Thus. D 54 (1996) 3915 [hep-th/9602135]. I.R.R. Witten. Math. [3] E. Witten. [14] S. Rept. Gubser and I. Strassler. Klebanov and A. B 435 (1998) 337 [hep-th/9805131]. Nekrasov.S. A.S. N. [12] O.R. Tseytlin and E. Klebanov and M. Maldacena. Klebanov. B 574 (2000) 263 [hep-th/9911096].23) Thus. Rev. Vafa. [4] S. Klebanov and N. the parameter ε2 which enters the deformed conifold equation has a dual interpretation as the gluino condensate6 . Phys. Nucl.R. B 578 (2000) 123 [hep-th/0002159]. [10] I. Theor. Nucl.R. D 58 (1998) 125025 [hep-th/9808075]. Gubser. H. Silverstein. Ooguri and Y. the holographic calculation tells us that λλ ∼ M ε2 · (α )3 (6. Phys. [8] S. [2] S.W. Phys. [11] I. . Klebanov. Klebanov and A. Nucl. TASI Lectures: Introduction to the AdS/CFT Correspondence [hep-th/0009139]. B 428 (1998) 105 [hep-th/9802109]. M. Phys. [6] I. Gubser. Adv. Adv. I. J. Theor. S. Phys. B 536 (1998) 199 [hep-th/9807080]. Kehagias.S.R. Nekrasov. Polyakov. Rev. I. Witten for collaboration on parts of the material reviewed in these notes and for useful input.S. Math. Theor. Klebanov and E. Nucl. Maldacena. Phys. Lett. 80 (1998) 4855 [hep-th/9802183]. Rev.R. Phys. 3 (1999) 1 [hep-th/9810201]. r3 (6. Lett.Phys. N. Strassler. Gubser. 323 (2000) 183 [hep-th/9905111].R. These are precisely the properties of λλ.

Marotta. Nucl. Dall’Agata. Rev. M. Nucl. Phys. Green. Vainshtein. More Anomalies from Fractional Branes [hep-th/0202195].L. B 435 (1995) 129 [hep-th/9411149]. Stable Non-BPS States and Their Holographic Duals [hep-th/0011185]. Nucl. Bardeen.S.P. Danielsson. Quant. Gubser. Bachas.A. Polchinski.R. Giddings. Witten. Marcus. [35] A. I. Klebanov. M. 2 (1985) L11. JHEP 007 (1997) 004 [hep-th/9612126]. D 59 (1999) 065011 [hep-th/9809022]. Seiberg. Romans. A. Argyres. L. Nucl. Adler and W. C.I. Phys. Rev. B 621 (2002) 157 [hep-th/0107057]. JHEP 9707 (1997) 002 [hep-th/9705074]. [28] E. C. B 556 (1999) 89 [hep-th/9905104]. D 32 (1985) 389. [26] C. Witten. Kachru and E.J. Aharony. Silverstein. Shifman and A. Jatkar and S. Lerda and R. Di Vecchia. [18] P.A. [17] L. M. [21] I. Phys. Witten. Lett. Pando Zayas and A. Phys. Ferretti and I. B 460 (1999) 281 [hep-th/9904187]. Ouyang and E. [19] A. [36] L. B 600 (2001) 219 [hep-th/0011146]. Susskind and E. JHEP 0104 (2001) 033 [hep-th/0102105].G.B. M. [41] S. Novikov. JHEP 0105 (2001) 028 [hep-th/0102172]. [34] J. Peet and J. D 58 (1998) 106002 [hep-th/9805129]. Klebanov. B 277 (1986) 456. Rev. D-branes. Klebanov and A. [45] O. Rev. [29] M.A.I. D 59 (1999) 025006 [hep-th/9807164].R. D 61 (2000) 066001 [hep-th/9905226]. Nucl. Nucl. Candelas and X. Suryanarayana. Mukhi and N. Randjbar-Daemi. Ooguri. Phys. [30] R. 182 (1969) 1517. Phys. Int. Lerda and R. 79 (1997) 1984 [hep-th/9705084]. Grav. Phys. Marotta.R. L. S. [24] E. Phys. Bertolini. Phys. Phys. Phys. Rev. P. Phys. [44] M. The holographic bound in anti-de Sitter space [hep-th/9805114]. Phys.I. Klebanov and P. B 153 (1985) 392. Nucl. Herzog. Gubser. Gimon and J. Frau. Gross and H. Remarks on the warped deformed conifold [hep-th/0108101].C.P. Frau. [40] N. Class. B 488 (1997) 159 [hep-th/9610205]. [42] I. Phys. Ferrara. quivers. Shifman. Lett. Phys. Moore. Rev.P. [32] S. [33] A. Douglas and M. B 572 (2000) 499 [hep-th/9911042]. [22] H. I. Phys. R. Nucl. A. Phys.R. Zakharov. and ALE instantons [hep-th/9603167]. [27] U. S. Buchel.R.J.R. B 342 (1990) 246. D 54 (1996) 1667 [hep-th/9601038]. [38] V. Herzog.W. Bertolini.R. P. Phys. Rev. Rev. . Nucl. van Nieuwenhuizen. A 16 (2001) 707 [hep-th/0011193]. M.R. Phys. Romans and P. [25] D. J. Klebanov et al. D 65 (2002) 105007 [hep-th/0202056]. Phys. D’Auria and S. Mod. Tsimpis.J. Phys. [39] P.A. [20] D. Kim. G. Witten. Polchinski.R. [31] C. A. Ceresole. Di Vecchia.: N = 1 Gauge/Gravity Dualities 421 [15] S. Kachru and J. u [23] S. Buchel. Tseytlin. [37] M. Douglas. JHEP 9807 (1998) 006 [hep-th/9805112]. [16] M. Lecture Notes. P. de la Ossa. Ouyang. S. Rev.A. Polchinski. Klebanov. Hierarchies from ﬂuxes in string compactiﬁcations [hep-th/0105097]. Klebanov and E. Lett. Douglas and G. Vainshtein and V. I. G¨naydin and N. Polchinski. Minasian and D. Tseytlin. B 229 (1983) 381.S. A. G. [43] M. Herzog.

P. Witten. Maldacena. M. Guijosa. N. K.J.S. B 555 (1999) 183 [hep-th/9902197]. [49] M. [72] F. [63] A. Girardello and A. Bachas. Tafjord. Skenderis. Pilch. B 477 (1996) 155 [hep-th/9605053]. Rev. JHEP 0011 (2000) 028 [hep-th/0010088]. Rev. Vafa. paper to appear. Phys. Vicari. Phys. Phys. D. B 500 (1997) 3 [hep-th/9703166]. [57] J. [65] K. Nucl. Gubser. Nucl. Supersymmetry and F-theory realization of the deformed conifold with 3-form ﬂux [hep-th/0010010]. [52] L. B 513 (1998) 87 [hep-th/9707244]. Herzog and I. L¨ and C. Callan. B 526 (2002) 388 [hep-th/0111078]. Freedman and K. Pope. JHEP 0201 (2002) 031 [hep-th/0110050]. Nucl. B. Phys. Phys. Savvidy and O. [51] K. Yankielowicz. Zaﬀaroni.J. Klebanov. O. Lett. c u M. Rev.N. Gibbons. [58] C. Theor. D 64 105019 (2001) 105019 [hep-lat/0107007]. Ricci-ﬂat Metrics. . Math. J. JHEP 0012 (2000) 014 [hep-th/0010048]. Loewy and J.R.W. K. Lett. DeWolfe. D. Pando Zayas and A. C 22 (2001) 379 [hep-th/9803001]. Dasgupta. 80 (1998) 4859 [hep-th/9803002]. Phys. B 600 (2001) 103 [hep-th/0011023]. Rev. B 396 (1997) 64 [hep-th/9612128]. Lett. Teper. Holographic Renormalization [hep-th/0112119].R. Lett. Brandhuber and K. [47] A. Phys. 86 (2001) 588 [hep-th/0008001]. Pope. Schweigert. Panagopoulos.Z. Tseytlin. [66] C. B 598 (2001) 530 [hep-th/0011041]. Bianchi. Harmonic c u Forms and Brane Resolutions [hep-th/0012011]. Phys. L¨ and C. Bianchi. Becker. [64] E. [71] A. JHEP 0002 (2000) 023 [hep-th/9912266]. Nucl. Adv. H. Yokono. Sfetsos. Henningson and K.Z. Eur. J. A. Sonnenschein. Hori. G. Maldacena and C. 1 (1998) 1 [hep-th/9706082]. Phys. Skenderis. Phys. Zaﬀaroni. Oz. [54] S. H. H. A. Krasnitz. Witten. Shenker. Tatar. Phys. Ooguri and Y. B 410 (1997) 27 [hep-th/9706127].h. L. E. Phys. [60] J. [55] K. Rey and J.G. [53] M.N. D 63 (2001) 026001 [hep-th/0009211]. JHEP 0101 (2001) 021 [hep-th/0009156]. Ohta and T. Douglas and C. Nunez. [48] M. Bigazzi.G. [62] M. Park and R. JHEP 0005 (2000) 048 [hep-th/0003037]. Del Debbio. Polchinski. Nucl. [67] C. k-string tensions in SU (N ) gauge theories [hep-th/0106185]. Freedman and K.H. V. [50] M. Superstrings and Topological Strings at Large N [hep-th/0008142]. [56] M. Sonnenschein and S. Itzhaki. Rossi and E. Lucini and M. B 507 (1997) 658 [hep-th/9706109]. Lett. [61] S.422 Unity from Duality: Gravity. Kaplunovsky. Shifman. Becker and M. Douglas and S. Nucl. [59] K. Strassler and A. Grana and J. Yee. JHEP 9807 023 (1998) [hep-th/9806087]. Phys. Phys. H. JHEP 0108 (2001) 007 [hep-th/0103163]. Oh. Nucl. Nucl. P. [68] C. Cvetiˇ. Cvetiˇ. Phys. B 447 (1995) 271 [hep-th/9503163]. Phys. Dvali and M. Phys. [70] G. [69] L. Brandhuber. J. Gauge Theory and Strings [46] M. Hanany. M.

S. Harvard University.LECTURE 6 DE SITTER SPACE A. .A. Cambridge. STROMINGER Department of Physics. MA 02138. U.

. . . . . . . . .1 Green functions and vacua . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 447 A Calculation of the Brown-York stress tensor 451 . . . . . . . . . . . .1 Asymptotic symmetries . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 Temperature . . . . . . . . . 437 3. . . . . . . . . . 435 2. . 428 2. 446 4. . . . . . . . . . . . .Contents 1 Introduction 425 2 Classical geometry of de Sitter space 427 2. . . . . . . . . 440 3.2 Schwarzschild-de Sitter . . . . . . . . . . . . . . . . . . . .1 Coordinate systems and Penrose diagrams .2 De Sitter boundary conditions and the conformal group . . . . . . . . . . 443 4 Quantum gravity in de Sitter space 446 4. . . . . . . . 436 3 Quantum ﬁeld theory on de Sitter space 437 3. . . . . . . . . . . . . . . . . .3 Entropy . . . . . .3 Geodesics . . . . . . . .

Springer-Verlag 2002 .2 . which says that the entropy S associated with an event horizon is its area A divided by 4G. A. MA. which is central to the dS/CFT correspondence. Bousso and A.3 Abstract These lectures present an elementary discussion of some background material relevant to the problem of de Sitter quantum gravity. Herzog and L. McAllister for comments on the manuscript. It describes how properties of event horizons in general relativity change as their parameters are We would like to thank C. especially the temperature and entropy of de Sitter space.S.D. This work was supported in part by DOE grant DE-FG02-91ER40655. Princeton NJ. Russia c EDP Sciences. Spradlin1. Cambridge. The ﬁrst two lectures discuss the classical geometry of de Sitter space and properties of quantum ﬁeld theory on de Sitter space. Nekrasov for organizing a very pleasant and productive summer school and for arranging ﬁnancial support. This is a macroscopic formula. 1 Harvard University. Landau Institute for Theoretical Physics. Douglas. and A. Maloney for useful discussions. It should be viewed in the same light as the macroscopic thermodynamic formulae that were ﬁrst studied in the 18th and 19th centuries. Strominger1 and A. The ﬁnal lecture contains a pedagogical discussion of the appearance of the conformal group as an asymptotic symmetry group. We are grateful to R. where G is Newton’s constant [1. 1 Introduction We begin these lectures with one of our favorite equations S= A · 4G (1.DE SITTER SPACE M. A. Bachas. M. A (previously lacking) derivation of asymptotically de Sitter boundary conditions is also given. F.A.A. is also supported by INTASOPEN-97-1312. Moscow.S. U. U. and to C. M. 3 L. David. 2].1) This is the Bekenstein-Hawking area-entropy law. is also supported by DOE grant DE-FG02-91ER40671. Bilal.S. Volovich1 .V. and N. 2 Princeton University.

Nevertheless the progress towards a complete understanding of (1. This behavior can be succinctly summarized by ascribing to them an entropy given by (1. On the other hand in de Sitter space the event horizon is observer dependent. Since then we’ve managed to creep away from the supersymmetric limit a little bit. It also applies to cosmological horizons. and it is diﬃcult even to see where the quantum microstates that we would like to count are supposed to be. Gauge Theory and Strings varied. In some stringy cases this agreement has been achieved. and indeed the ﬁrst black holes whose entropy was counted microscopically were supersymmetric.1) that we have managed to understand has led to all kinds of interesting insights. such as the horizon in de Sitter space. One of the surprising and impressive features of this formula is its universality. you would be able to count those microstates and compare your result to the Bekenstein-Hawking formula and see that they agree.1) was not given at the time that the law was discovered in the early 70s. because we only understand special kinds of black holes–among which Schwarzschild black holes are not included–and we certainly don’t understand much about cosmological event horizons. So supersymmetry is a crutch that we will need to throw away before we can do anything about de Sitter space.1) is still very limited. Indeed there is a very simple . Why has there been signiﬁcant progress in understanding black hole entropy. but almost no progress in understanding the entropy of de Sitter space? One reason is that one of the principal tools we’ve used for understanding black hole entropy is supersymmetry.1) in very special cases of black holes which can be embedded into string theory [4]. After Boltzmann’s work we tend to think of entropy in microscopic statistical terms as something which counts the number of microstates of a system.1). shapes and rotation. but not very far. There has been some deﬁnite but still limited progress in understanding the microscopic origin of (1. Black holes can be supersymmetric. Then if you could provide the correct description of that localized object. That little piece of (1. A complete understanding of this law. as well as to black strings and to all of the strange new objects we’ve found in string theory. ultimately culminating in the AdS/CFT correspondence [5].426 Unity from Duality: Gravity. Such an interpretation for the entropy (1. like the event horizon in de Sitter space [3]. In some ways cosmological horizons are much more puzzling than black hole horizons because in the black hole case one may expect that the black hole is a localized object with some quantum microstates. It applies to all kinds of black holes with all kinds of charges. and certainly we never managed to get all the way to Schwarzschild black holes. and in particular of the statistical origin of this law. is undoubtedly one of the main keys to understanding what quantum gravity is and what the new notions are that replace space and time in quantum gravity.

1 See. as shown in Figure 1. The classical geometry of de Sitter space is described in Section 2.2) however [7]. While the importance of understanding de Sitter quantum gravity has been evident for decades. where is a parameter with units of length called the de Sitter radius. which is the isometry group of dS d . which leads to the dS/CFT correspondence [27]. So perhaps de Sitter quantum gravity is a nut ready to be cracked. These lectures are mostly an elementary discussion of the background material relevant to the problem of de Sitter quantum gravity.1) in ﬂat d+1-dimensional Minkowski space Md. A pedagogical derivation is given of the appearance of the the two dimensional conformal group in three dimensional de Sitter space. of the asymptotically de Sitter boundary conditions on the metric. (2. in Section 4 we turn to some recent work on de Sitter quantum gravity. it has recently been receiving more attention [7– 42]. One reason for this is the recent astronomical observations which indicate that the cosmological constant in our universe is positive [43–46]. This hypersurface in ﬂat Minkowski space is a hyperboloid. related. is manifest. The embedding (2. The de Sitter metric is the induced metric from the standard ﬂat metric on Md. 2 Classical geometry of de Sitter space The d-dimensional de Sitter space dS d may be realized as the hypersurface described by the equation 2 2 2 −X0 + X1 + · · · + Xd = 2 (2. This section also contains a derivation. Scalar quantum ﬁeld theory in a ﬁxed de Sitter background is in Section 3.1 .1) is a nice way of describing de Sitter space because the O(d. 1) symmetry. The Appendix contains a calculation of the asymptotic form of the Brown-York stress tensor. and the Einstein tensor satisﬁes Gab + Λgab = 0. Furthermore one can show that dS d is an Einstein manifold with positive scalar curvature. missing in previous treatments. A second. obstacle to progress in understanding de Sitter space is that so far we have not been able to embed it in a fully satisfactory manner into string theory.: De Sitter Space 427 observation [6] that de Sitter space is inconsistent with supersymmetry in the sense that there is no supergroup that includes the isometries of de Sitter space and has unitary representations1. Strominger et al. Finally. . A second reason is that recent progress in string theory and black holes provides new tools and suggests potentially fruitful new angles.A.1 .

We will frequently make use of coordinates on the sphere S d−1 . sin θ1 · · · sin θd−2 sin θd−1 . sin θ1 cos θ2 . Then it is clear d that i=1 (ω i )2 = 1. sin θ1 · · · sin θd−2 cos θd−1 . but 0 ≤ θd−1 < 2π. 1. . where (d − 2)(d − 1) 2 2 is the cosmological constant. The dotted line represents an extremal volume S d−1 . which is conveniently parametrized by setting ω1 ω2 ω d−1 ωd = = .3) = 1. . (2. θi ).4) where 0 ≤ θi < π for 1 ≤ i < d − 1. This coordinate system is obtained by setting X0 = sinh τ. and the metric on S d−1 is d dΩ2 = d−1 i=1 2 2 2 (dω i )2 = dθ1 + sin2 θ1 dθ2 + · · · + sin2 θ1 · · · sin2 θd−2 dθd−1 . Henceforth we will set Λ= 2. Hyperboloid illustrating de Sitter space. . = = cos θ1 . Global coordinates (τ.428 Unity from Duality: Gravity. (2.1 Coordinate systems and Penrose diagrams We will now discuss a number of coordinate systems on dS d which give diﬀerent insights into the structure of dS d .5) a. Gauge Theory and Strings X 0 Fig. (2.

They are the surfaces where all null geodesics originate and terminate.1) for any point (τ. cos T (2. b. These coordinates are related to the global coordinates by cosh τ = 1 .9) so that we have −π/2 < T < π/2. If a geodesic is null with respect to the metric (2.8) In these coordinates dS d looks like a d−1-sphere which starts out inﬁnitely large at τ = −∞.11) So from the point of view of analyzing what null geodesics do in dS d we are free to work with the metric (2. then shrinks to a minimal ﬁnite size at τ = 0. ds2 = −dτ 2 + (cosh2 τ )dΩ2 . (2. then it is also null with respect to the conformally related metric d˜2 = (cos2 T )ds2 = −dT 2 + dΩ2 . . θi ). Strominger et al.10).4).A.10). s d−1 (2. . which looks simpler than (2.: De Sitter Space Xi = ω i cosh τ.1 2 2 2 ds2 = −dX0 + dX1 + · · · + dXd .6) we obtain the induced metric on dS d . 429 (2. d−1 (2. ωi ). Light rays travel at 45◦ angles in this diagram.11).6) where −∞ < τ < ∞ and the ω i are as in (2. The metric in these coordinates takes the form ds2 = 1 (−dT 2 + dΩ2 ). Note that a light ray which starts at the north pole at I − will exactly reach the south pole by the time it reaches I + inﬁnitely far in the future. then grows again to inﬁnite size as τ → +∞. . d. The Penrose diagram 2 contains all the information about the causal structure of dS d although distances are highly distorted.7) plugging in (2. while timelike surfaces are more vertical than horizontal and spacelike surfaces are more horizontal than vertical. . d−1 cos2 T (2. From the ﬂat metric on Md. i = 1. In this diagram each point is actually an S d−2 except for points on the left or right sides. Conformal coordinates (T. which lie on the north or south pole respectively. It is not hard to check that these satisfy (2. I + are called past and future null inﬁnity. .10) This is a particularly useful coordinate system because it enables us to understand the causal structure of de Sitter space. The surfaces marked I − .

One of the peculiar features of de Sitter space is that no single observer can access the entire spacetime. 2. while every point in the interior represents an S d−2 . This is qualitatively diﬀerent from Minkowski space. where a timelike observer will eventually have the entire history of the universe in his/her past light cone. The north and south poles are timelike lines. A classical observer sitting on the south pole will never be able to observe anything past the diagonal line stretching from the north pole at I − to the south pole at I + .430 Unity from Duality: Gravity. I South Pole North Pole North Pole I South Pole O O I I Fig. These diagrams show the regions O− and O+ corresponding respectively to the causal past and future of an observer at the south pole. We often get into trouble in physics when we try to describe more than we are allowed to observe–position and momentum in quantum mechanics. for example. Therefore in attempting to develop de Sitter quantum gravity we should be aware of what can and cannot be observed. which is the only part of de Sitter space that an observer on the south pole will ever be able to send . for example. Also shown in Figure 3 is the region O+ . Gauge Theory and Strings I South Pole North Pole I Fig. The dashed lines are the past and future horizons of an observer at the south pole. A horizontal slice is an S d−1 . Penrose diagram for dS d . The conformal time coordinate T runs from −π/2 at I − to +π/2 at I + . 3. This region is marked as O− in Figure 3.

The surfaces of constant t are spatial sections of de Sitter space which are inﬁnite volume d−1-planes with the ﬂat metric. . The dashed lines are slices of constant t in planar coordinates. Note that each slice is an inﬁnite ﬂat d−1-dimensional plane which extends all the way down to I − . The intersection O+ ∩ O− is called the (southern) causal diamond. 1 cosh t − xi xi e−t . . i = 1. Strominger et al. I South Pole North Pole O I Fig. . For example if she/he wishes to know the weather anywhere in the southern diamond. . Planar coordinates (t. . From the diagram it . to which a query can never be sent. . . d − 1. or the upper diamond of O+ . but only the region O− and are therefore appropriate for an observer on the south pole. . To deﬁne this coordinate system we take X0 Xi X d = = = 1 sinh t − xi xi e−t .A.12) The metric then takes the form ds2 = −dt2 + e−2t dxi dxi . The northern diamond on the left of 3 is completely inaccessible to an observer on the south pole. 4. 2 (2.: De Sitter Space 431 a message to. xi ). i = 1. 2 xi e−t . It is this region that is fully accessible to the observer on the south pole. (2. c. d − 1. This is not possible in the lower diamond of O− .13) These coordinates do not cover all of de Sitter space. from which a response cannot be received. The slices of constant t are illustrated in Figure 4. a query can be sent to the appropriately located weather station and the response received before I + is reached.

2 cosh t. so that it becomes null.16) . = rω a .14) then the metric takes the form ds2 = −(1 − r2 )dt2 + dr2 + r2 dΩ2 . not S d−1 .15) In this coordinate system ∂/∂t is a Killing vector and generates the symmetry t → t + constant. a = 1. θa ). and you can get there along a spatial trajectory from anywhere in O− . This coordinate system is the de Sitter analog of the (outgoing) Eddington-Finkelstein coordinates for a Schwarzschild black hole. One of the reasons to want a timelike Killing vector is so that we can use it to deﬁne time evolution. while in the northern diamond the vector is pointing towards the past! Thus ∂/∂t in static coordinates can only be used to deﬁne a sensible time evolution in the southern diamond of de Sitter space. d. . The t in these coordinates is not the same as the t in planar coordinates. Eddington-Finkelstein coordinates (x+ . In the top and bottom diamonds. . But from (2. . d−2 1 − r2 (2. The absence of a globally timelike Killing vector in de Sitter space has important implications for the quantum theory. we deﬁne x+ by the equation dx+ = dt + dr . If we write X0 Xa Xd = 1 − r2 sinh t. Gauge Theory and Strings is clear that every surface of constant t intersects I − at the north pole. In these coordinates the time t is not a Killing vector. . 1 − r2 (2.432 Unity from Duality: Gravity. d − 2. but we are running out of letters! Note also that for these coordinates and the following ones we will need a parametrization of S d−2 . This coordinate system is constructed to have an explicit timelike Killing symmetry. This happens because I − is very large. It may seem puzzling–and is certainly one of the salient features of de Sitter space–that a spatial plane can make it to the inﬁnite past. e. . In Figure 5 we illustrate what the Killing vector ﬁeld ∂/∂t is doing when extended to the various diamonds of the Penrose diagram. d − 1. or in other words to deﬁne the Hamiltonian. . . The horizons are at r2 = 1. ∂/∂t is spacelike. r. a = 1. θa ). and the southern causal diamond has 0 ≤ r ≤ 1. with the south pole at r = 0. Static coordinates (t. . Starting from the static coordinates. r. = 1−r (2. and the only manifest symmetries are translations and rotations of the xi coordinates.15) we see that at r = 1 the norm of ∂/∂t vanishes.

These coordinates cover the causal past O− of an observer at the south pole while still keeping the symmetry manifest. Finally we take x− = ln U.17) The same symmetries are manifest in this coordinate system as in the static coordinates since ∂/∂t at ﬁxed r is the same as ∂/∂x+ at ﬁxed r.20) −x− · 2 f. which we can solve to obtain x+ = t + In these coordinates the metric is ds2 = −(1 − r2 )(dx+ )2 + 2dx+ dr + r2 dΩ2 . d−2 where r = tanh x + 1 1+r ln . x+ = − ln(−V ).21) . Strominger et al.19) (2. (2.18) 1 1+r ln · 2 1−r (2. Past and future null inﬁnity I ± are at r = ∞. This Penrose diagram shows the direction of the ﬂow generated by the Killing vector ∂/∂t in static coordinates. 5.A. 2 1−r (2. Lines of constant x+ are the null lines connecting I − with the south pole depicted in Figure 3. and the southern causal diamond is the region with 0 ≤ r ≤ 1 on the right hand side. Kruskal coordinates (U. V. x− ))dx+ dx− + r2 dΩ2 . d−2 (2. θa ). We can also deﬁne x− = t − so that the metric takes the form ds2 = −(1 − r2 (x+ . The horizons (dotted lines) are at r 2 = 1.: De Sitter Space 433 Fig.

Global coordinates foliate de Sitter space with spheres.22) These coordinates cover all of de Sitter space. ω a sinh τ sinh ψ.24) (2. . . The arrows denote the directions of increasing U and V .25) in which the metric takes the form ds2 = −dτ 2 + sinh2 τ (dψ 2 + sinh2 ψ dΩ2 ). . the horizons correspond to U V = 0. and U V = 1 are I + and I − . d−2 (1 − U V )2 (2. d−2 and surfaces of constant τ are d−1-dimensional hyperbolic planes. cosh τ . . One can also foliate de Sitter space with spaces of constant negative curvature by using the hyperbolic coordinates X0 Xa Xd = = = sinh τ cosh ψ. V UV=1 U UV=1 V= 0 UV=1 Fig. and planar coordinates foliate de Sitter space with planes. Gauge Theory and Strings in which case r= Then the metric takes the form ds2 = 1 (−4dU dV + (1 + U V )2 dΩ2 ). Find X 0 . and I ± correspond to U V = 1. Hyperbolic coordinates (τ . In this Penrose diagram the coordinate axes U = 0 and V = 0 are the horizons. The southern diamond is the region with U > 0 and V < 0. and θa for the Kruskal coordinates. Exercise 1.434 Unity from Duality: Gravity. U V = −1 are the north and south poles. g. 6. V . . X d as functions of U. θa ). UV=1 U =0 (2. ψ. The north and south poles correspond to U V = −1. The Kruskal coordinate system covers all of de Sitter space.23) 1 + UV · 1 − UV (2.

rH (2. We can learn a little more about the solution (2. (2. at rH = 1 − 8GE. and as E goes to zero this reduces to the usual horizon in empty de Sitter space. The fact that there is only a de Sitter horizon and not a black hole horizon is not surprising in light of the fact that in three dimensional ﬂat space there are no black holes. In d dimensions in static coordinates the SdS d metric takes the form ds2 = − 1 − 2m − r2 rd−3 dt2 + 1 1− 2m r d−3 − r2 dr2 + r2 dΩ2 . φ = rH φ. In three dimensions there is only one horizon. (2.A. This solution represents a black hole in de Sitter space.2 Schwarzschild-de Sitter 435 The simplest generalization of the de Sitter space solution is Schwarzschildde Sitter.27) by looking near r = 0.28) is simply ds2 = −dt 2 + dr 2 + r 2 dφ 2 . Note that the two horizons approach each other as m is increased.27) where we have normalized the energy E of the Schwarzschild black hole appropriately for√ three dimensions. which we abbreviate as SdS. One reason to introduce SdS is that it plays an important role in the work of Gibbons & Hawking [3] determining the entropy of pure de Sitter space. r = r/rH . In general there are two horizons (recall that these are places where the timelike Killing vector ∂/∂t becomes null). so that there is a maximum size black hole which can ﬁt inside de Sitter space before the black hole horizon hits the de Sitter horizon. one of which is the black hole horizon and the other of which is the de Sitter horizon. Strominger et al. (1 − 8GE − r2 ) (2.: De Sitter Space 2.28) Now we can rescale the coordinates by deﬁning t = rH t. which will be reviewed in Section 3. d−2 (2. For this purpose it will be convenient to focus on the three dimensional Schwarzschild-de Sitter solution [49] ds2 = −(1 − 8GE − r2 )dt2 + dr2 + r2 dφ2 .26) where m is a parameter related to the black hole mass (up to some d-dependent normalization constant).3. where ds2 behaves like 2 ds2 ∼ −rH dt2 + dr2 2 2 2 + r dφ .30) .29) In the rescaled coordinates the metric (2.

i. X ) between X and X by P = cos D. 2. . Therefore there is a conical singularity with a positive deﬁcit angle at the origin. we note a few important properties of P for later use. φ is identiﬁed modulo 2πrH . but it is not quite ﬂat space because while φ was identiﬁed modulo 2π. . If the solution is maximally extended one ﬁnds there is also point-like mass of the same size at the north pole [49]. Show that SdS 3 is a global identiﬁcation of dS 3 . rather than a black hole. y i ) = cosh(t − t ) − e−t−t δij (xi − y i )(xj − y j ). Gauge Theory and Strings This looks like ﬂat space. xi . Let us think of the sphere as being embedded in ﬂat Euclidean space. but we can do something similar for dS d . It is useful to deﬁne a quantity P by R2 P (X. with the embedding equation δij X i X j = R2 . .27) as a pointlike mass. 1) (2. the geodesic separating X and X is spacelike for P < 1 and timelike for P > 1. There we can deﬁne P (X. then the geodesic distance is equal to zero. ηij = diag(−1.3 Geodesics Our last topic in the classical geometry of de Sitter space is geodesics. at the south pole of dS 3 . You may be familiar with the fact that if you put a point-like mass in ﬂat three dimensional Minkowski space you would also get a conical deﬁcit angle at the location of the particle. It is a little harder to visualize. . We can easily write explicit formulas for P (X. If P = 1. this is related to the geodesic distance D(X. In general. then there is only one independent SO(n + 1)-invariant quantity that we can associate to the two points. so the two points X and X coincide or are separated by a null geodesic. while for P < −1 the geodesic between X and the antipodal point of X is timelike. For example. X ) ≡ δij X i X j = R2 cos θ. n + 1. To conclude. . which are related by D = Rθ. . 1. 2 (2.32) The expression for P is simple in terms of the X’s but can get complicated when written in a particular coordinate system. . Hence we recognize (2. This quantity P will turn out to be a more convenient invariant to associate to two points in de Sitter space.31) (recall that we have set the de Sitter radius to one). . We can also consider taking antipodal points X = −X. or equivalently the angle θ between them. t . in which case P = −1. That is the geodesic distance D. X ) = ηij X i X j . in planar coordinates we have 1 P (t. In general P = −1 when the antipodal point of X lies on the light cone of X .436 Unity from Duality: Gravity. X ) in the various coordinate systems discussed above. Exercise 2. . For points in a common causal diamond. j = 1. It is clear that if we take two points on the sphere S n of radius R.

Then G(X. (3.1) in de Sitter space will require an understanding of quantum gravity on de Sitter space. as discussed below. (3. Y ) = G(P (X.1 Green functions and vacua Let us consider a scalar ﬁeld in dS d with the action S=− 1 2 √ dd x −g (∇φ)2 + m2 φ2 . which obeys the free ﬁeld equation (∇2 − m2 )G(X. Y ) between X and Y 2 . 2 (3. and/or by multiplying by a step function in time. Y )). There are other two point functions that one can discuss: retarded. X) is insensitive to the time ordering between points. Hadamard and so on. In this section we will take a baby step in that direction by considering a single free massive scalar ﬁeld on a ﬁxed background de Sitter spacetime. 1)) invariant. 50–60].A. This turns out to be a very rich subject which has been studied by many authors [3.3) (3. which is SO(d. Writing G(X.4) With the change of variable z = equation z(1 − z)G + 1+P 2 this becomes a hypergeometric d − dz G − m2 G = 0. 3.: De Sitter Space 3 Quantum ﬁeld theory on de Sitter space 437 Ultimately. Y ) = 0. for example by taking the real or imaginary part. Let us assume that the state |0 in (3. We will study the Wightman function G(X. a complete understanding of the entropy-area relation (1. Feynman.1) Since this is a free ﬁeld theory. but these can all be obtained from the Wightman function (3. Strominger et al. can not be written as a function of P alone. Y ) will be de Sitter invariant. Because of this the i prescription for G. (3. 1) (but not O(d. advanced. Y ) = 0|φ(X)φ(Y )|0 . and so at generic points can only depend on the de Sitter invariant length P (X. all information is encoded in the two-point function of φ. 1) de Sitter group.2). Y ) = P (Y.3) reduces to a diﬀerential equation in one variable P 2 (1 − P 2 )∂P G − dP ∂P G − m2 G = 0.2) is invariant under the SO(d. (3. .2) where ∇2 is the Laplacian on dS d .5) 2 P (X.

d F d h+ . Γ( d ) 2 Comparing with the usual short-distance singularity 2(d−2)πd/2 (D2 )1−d/2 ﬁxes the coeﬃcient to be cm.8) where D = cos−1 P is the geodesic separation between the two points. but we should recall that antipodal points in de Sitter space are always separated by a horizon. . The equation (3. Gauge Theory and Strings whose solution is G = cm. (3.438 Unity from Duality: Gravity. . namely replacing X 0 −Y 0 with X 0 −Y 0 −i . The second linearly independent solution to (3. h− . so if G(P ) is a solution then G(−P ) is also a solution.7) The hypergeometric function (3.4) is therefore F d 1−P h+ .9) The prescription for going around the singularity in the complex plane is also the same as in Minkowski space. 2 2 ∼ D2 4 1−d/2 Γ( d )Γ( d − 1) 2 2 .d = 41−d/2 Γ( d ) Γ(h+ )Γ(h− ) Γ(h+ )Γ(h− ) 2 × · = d d d/2 Γ( 2 )Γ( 2 − 1) 2(d − 2)π (4π)d/2 Γ( d ) 2 (3. Γ(h+ )Γ(h− ) (3.6) where cm. At short distances the scalar ﬁeld is insensitive to the fact that it is in de Sitter space and the form of the singularity is precisely the same as that of the propagator in ﬂat d-dimensional Minkowski space.10) can be thought of as arising from an image source behind the horizon. Near z = 1 the hypergeometric function behaves as F d 1+P h+ .d . This singularity sounds rather unphysical at ﬁrst. h− .10) is nonsingular everywhere within an observer’s horizon.4) clearly has a P → −P symmetry. 2 (3.d is a normalization constant to be ﬁxed shortly. which corresponds to X being null separated from the antipodal point to Y . and a branch cut for 1 < P < ∞. 2 2 · (3. and (3. and h± = 1 (d − 1) ± 2 (d − 1)2 − 4m2 . We can use this fact to ﬁx the normalization constant cm. The singularity occurs when the points X and Y are separated by a null geodesic. or P = 1. z . h− . .6) has a singularity at z = 1. Hence the “unphysical” singularity can not be detected by any experiment.10) The singularity is now at P = −1. . The Green function (3.

which modes do we associate with creation operators in (3. (3. which when analytically continued back to de Sitter space yields (3.15) where the integral is taken over a complete spherical spacelike slice in dS d and the result is independent of the choice of this slice.12) and which do we associate with annihilation operators? In Minkowski space we take positive and negative frequency modes. On the sphere there is a unique Green function.6) and (3.11) That is. The question is. but are usually discarded as somehow “unphysical”. as we try to understand the quantum theory of de Sitter space these funny extra vacua will surely turn out to have some purpose in life. we write an expansion for the scalar ﬁeld in terms of creation and annihilation operators of the form φ(X) = k ak uk (X) + a† u∗ (X) . Corresponding to this one-parameter family of Green functions is a one-parameter family of de Sitter invariant vacuum states |α such that Gα (X.10). However. A vacuum state |0 is deﬁned as usual by saying that it is annihilated by all annihilation operators an |0 = 0. Let us say a few more words about the vacuum states.16) .14) (3.6). These vacua are discussed in detail in [58. (3. Strominger et al. a† ] = δkl .A. u∗ ∼ eiEt f ∗ (x).13) and are normalized with respect to the invariant Klein-Gordon inner product (uk . l ↔ (3.: De Sitter Space 439 De Sitter space therefore has a one parameter family of de Sitter invariant Green functions Gα corresponding to a linear combination of the solutions (3. De Sitter Green functions are often discussed in the context of analytic continuation to the Euclidean sphere. If we work in static coordinates and take t → iτ . l The modes uk (X) satisfy the wave equation (∇2 − m2 )uk = 0. k k (3. Y ) = α|φ(X)φ(Y )|α . (3. u ∼ e−iEt f (x). 59]. ul ) = −i dΣµ uk ∂ µ u∗ = δkl .12) where ak and a† satisfy k [ak . the dS d metric becomes the metric on the sphere S d .

(3. 59] that there is a family of such divisions. It turns out [51. Such a detector is called an Unruh detector and may be modeled by a coupling of the scalar ﬁeld φ(x(τ )) along the worldline x(τ ) of the observer to some operator m(τ ) acting on the internal detector states ∞ g −∞ dτ m(τ )φ(x(τ )). But if we had made some other choice of time coordinate then we would have a diﬀerent vacuum |0 . The resulting vacuum will then be de Sitter invariant. Gauge Theory and Strings respectively to multiply the annihilation and creation operators. Thus we will conclude that de Sitter space is naturally associated with a temperature [53]. i. we must be careful to give a coordinate invariant characterization of the temperature.17) where g is the strength of the coupling and τ is the proper time along the observer’s worldline. 58.18) .e. All observers will agree on whether or not the detector has binged. A good way to achieve this is to consider an observer equipped with a detector. The detector will have some internal energy states and can interact with the scalar ﬁeld by exchanging energy. Let H denote the detector Hamiltonian. Hence the question “How many particles are present?” is not well-deﬁned independently of a choice of coordinates. with energy eigenstates |Ej . although they may disagree on whether the bing was caused by particle emission or absorption. Since the notion of a particle is observer-dependent in a curved spacetime. 3. (3. The detector could for example be constructed so that it emits a “bing” whenever its internal energy state changes. But in a general curved spacetime there is no canonical choice of a time variable with respect to which one can classify modes as being positive or negative frequency. which we could express as a linear combination of the |n ’s.440 Unity from Duality: Gravity. and a corresponding family of Green functions such as Gα . H|Ej = Ej |Ej . we can get a vacuum state |0 and then the state (a† )n |0 ≡ |n is said to have n particles in it. If we make a choice of time coordinate. In order to preserve classical symmetries of dS d in the quantum theory. which we will calculate. This is an important and general feature of quantum ﬁeld theory in curved spacetime. 1). by emitting or absorbing scalar particles. we would like to ﬁnd a way to divide the modes into u and u∗ that is invariant under SO(d.2 Temperature In this section we will show that an observer moving along a timelike geodesic observes a thermal bath of particles when the scalar ﬁeld φ is in the vacuum state |0 .

22) Since we are only interested in the probability for the detector to make the transition from Ei to Ej .19) We will calculate the transition amplitude from a state |0 |Ei in the tensor product of the scalar ﬁeld and detector Hilbert spaces to the state Ej | β|. x(τ )) is the Green function (3. (3.20) Using m(τ ) = eiHτ m(0)e−iHτ . then P is given in static coordinates by P = cosh(τ − τ ).2).22) and sum over the ﬁnal state |β of the scalar ﬁeld. this can be written as ∞ (3.21) gmji −∞ dτ ei(Ej −Ei )τ β|φ(x(τ ))|0 · (3. which will not be measured.: De Sitter Space and let mij be the matrix elements of the operator m(τ ) at τ = 0: mij ≡ Ei |m(0)|Ej · 441 (3. Therefore everything inside the integral (3.A. x(τ )). We can divide out this factor and discuss the transition probability per unit proper time along the detector worldline. where β| is any state of the scalar ﬁeld.24) The ﬁrst hint that (3. To ﬁrst order in perturbation theory for small coupling g.23) where G(x(τ ). . we should square the amplitude (3. and if we consider for simplicity an observer sitting on the south pole. x(τ )).23) depends only on τ − τ and we get an inﬁnite factor from integrating over τ + τ . and Green functions which are periodic in imaginary time are thermal Green functions. the desired amplitude is ∞ g −∞ dτ Ej | β|m(τ )φ(x(τ ))|0 |Ei · (3. (3. Strominger et al.24) has something to do with a thermal response is that the function G is periodic in imaginary time under τ → τ + 2πi. which is then given by ˙ P (Ei → Ej ) = g 2 |mij |2 ∞ −∞ dτ e−i(Ej −Ei )τ G(cosh τ ). The Green function is a function only of the geodesic distance P (x(τ ). Using β |β β| = 1. we ﬁnd the probability P (Ei → Ej ) = g 2 |mij |2 ∞ −∞ dτ dτ e−i(Ej −Ei )(τ −τ ) G(x(τ ).

25) holds for the transition probabilities calculated in (3. The dotted lines signify the closure of the contour at inﬁnity. (3.442 Unity from Duality: Gravity. Since the total integral around this contour is zero. Let us now show that (3.28).27) which is the principle of detailed balance in a thermal ensemble. if the transition probabilities are related by (3. Then it is clear that the total transition rate R from Ei to Ej is the same as from Ej to Ei : ˙ R(Ei → Ej ) = N e−βEi P (Ei → Ej ) = R(Ej → Ei ). τ τ=0 C τ=−2πi Fig. (3.28) . Ni = N e−βEi .25) describes the transition probabilities of a system in a thermal bath of particles at temperature T = 1/β.24) has singularities in the complex τ -plane at τ = 2πin for any integer n. (3. The integrand in (3. Consider integrating the function e−i(Ej −Ei )τ G(cosh τ ) around the contour shown in Figure 7.25) and that the energy levels of the detector were thermally populated. then there is no change in the probability distribution for the energy levels with time. we have ∞ dτ −∞ −i(Ej −Ei )τ −∞−iβ G(cosh τ ) + +∞−iβ dτ e−i(Ej −Ei )τ G(cosh τ ) = 0.24).26) where N is some normalization factor. 7. Gauge Theory and Strings To investigate the nature of a thermal state.24) has singularities in the complex τ -plane at τ = 2πin for any integer n. The integrand in (3.26). In other words. So (3.25) and the population of the states is thermal as in (3. let us suppose it were true (as will be demonstrated shortly) that ˙ ˙ P (Ei → Ej ) = P (Ej → Ei )e−β(Ej −Ei ) . This ﬁgure shows the contour C used in the integral (3. so that (3.

30) to ﬁnd the entropy. The constant of integration is determined by requiring that a black hole of zero mass has zero entropy. 1) de Sitter isometry group. for ﬁxed coupling constant. but if you know it as a function of the black hole mass then you can simply integrate (3. So if you know the value of the temperature just for one value of EBH you will not be able to get the entropy. what is EdS in (3. Although we performed this calculation only for an observer stationary at the south pole. So for de Sitter space one would expect to use the relation dSdS 1 = dEdS TdS (3.29) by dimensional where we have restored the factor of the de Sitter radius analysis. Since the Green function used in this calculation is de Sitter invariant. The black hole entropy SBH can then be found by integrating the thermodynamic relation 1 dSBH = .3 Entropy In this subsection we will associate an entropy to de Sitter space. For the case of black holes one can use similar methods as those in the previous section to calculate the temperature TBH of the black hole.30) where EBH is the energy or mass of the black hole.A. 2π (3.1.: De Sitter Space 443 where β = 1/2π. Now redeﬁning the variable of integration in the second integral as τ = −τ − iβ we get precisely the desired relation (3. Strominger et al. all timelike geodesics in de Sitter space are related to each other by the SO(d. The contour in Figure 7 corresponds to the pole prescription for the Wightman function as discussed in Section 3. where the analysis simpliﬁes considerably.31)? One might try to vary the cosmological . the result for the temperature is the same for any observer moving along a timelike geodesic. We conclude that any geodesic observer in de Sitter space will feel that she/he is in a thermal bath of particles at a temperature TdS = 1 .25). The problem in de Sitter space is that once the coupling constant of the theory is chosen there is just one de Sitter solution. dEBH TBH (3.31) to ﬁnd the entropy SdS . 3. We will restrict our attention to dS 3 . In other words. whereas in the black hole case there is a whole one parameter family of solutions labeled by the mass of the black hole.

35) but instead dSSdS 1 · = d(−EdS ) TSdS (3.38) . 1 − 8GE = 2π one ﬁnds that the entropy is equal to SSdS = − AH · 4G (3. Let us instead follow Gibbons & Hawking [3] and use the one parameter family of Schwarzschild-de Sitter solutions to see how the temperature varies as a function of the parameter E labeling the mass of the black hole. The SdS 3 solution in static coordinates is ds2 = −(1 − 8GE − r2 )dt2 + dr2 + r2 dφ2 .444 Unity from Duality: Gravity. (1 − 8GE − r2 ) (3.36) This diﬀers by a minus sign from the famous formula (1. dE TSdS (3.35) and writing the result in terms of the area AH of the de Sitter horizon at √ rH = 1 − 8GE which is given by √ AH . but that is rather unphysical as it is the coupling constant. Gauge Theory and Strings constant. Exercise 3. Show that this Green function is periodic in imaginary time with periodicity τ →τ+√ 2πi · 1 − 8GE (3.1)! What did we do wrong? Gibbons and Hawking suggested that to get the de Sitter entropy we should use not (3.34) TSdS = 2π Using the formula 1 dSSdS = .33) From the exercise and the discussion in the previous section we conclude that the temperature associated with the Schwarzschild-de Sitter solution is √ 1 − 8GE · (3. One would be going from one theory to another instead of from one conﬁguration in the theory to another conﬁguration in the same theory.37) (3.32) Find a Green function for SdS 3 by analytic continuation from the smooth Euclidean solution.

I South Pole North Pole t= 0 I Fig. and hence vanishes on any closed surface. In (3. Now in general relativity the expression for the energy on a surface is the integral of a total derivative. which reduces to a surface integral on the boundary of the surface. Therefore the northern singularity of Schwarzschild-de Sitter behind the horizon actually carries negative energy. If we put something with positive energy on the south pole. we are forced to ascribe negative energy to the positive deﬁcit angle at the north pole. This can be seen quite explicitly in the Schwarzschild-de Sitter solution.38). Strominger et al. 8. then because the Killing vector ∂/∂t used to deﬁne the energy changes direction across the horizon. If we ascribe positive energy to a positive deﬁcit angle at the south pole. but if we more sensibly vary with respect to the energy at the north pole.37). Consider a closed surface in de Sitter space such as the one shown in Figure 8. With no black hole.: De Sitter Space 445 This looks funny but in fact there is a very good reason for using this new formula. is supposed to correspond to the entropy of the stuﬀ behind the horizon which we can’t observe. The de Sitter entropy.35) we varied with respect to the energy at the south pole. and ended up with the wrong sign in (3. The energy associated to the Killing vector ∂/∂t (indicated by the arrows) along the spacelike slice t = 0 (solid line) must vanish.A. but we saw in one of the exercises that in the SdS 3 solution there is a positive deﬁcit angle at both the north and south poles. although we don’t know exactly how to think about it. the spacelike slice in Figure 8 is an S 2 . If we ascribe positive energy to the positive deﬁcit angle at the south pole. then we should use the formula (3. then we must ascribe negative energy to a positive deﬁcit angle at the south pole since the Killing vector ∂/∂t runs in the opposite direction behind the horizon. then necessarily there will be some negative energy on the north pole. Then we arrive at the entropy for .

The tool which will allow us to reach this conclusion is an analysis of the asymptotic symmetry group for gravity in dS 3 . In fact the generator will be proportional to the charge operator.1) indeed applies to three dimensional Schwarzschild-de Sitter. in the sense that correlation functions of an operator φ inserted at points xi on I − or I + are generated by a two dimensional Euclidean CFT: φ(x1 ) · · · φ(xi ) dS 3 ↔ Oφ (x1 ) · · · Oφ (xi ) S2 . 26. Equation (4.1) expresses the dS/CFT correspondence. In this section we will give a pedagogical discussion of several aspects of some recent eﬀorts in this direction [27] (some similar ideas appeared in [7. A gauge transformation which goes to zero at spatial inﬁnity will annihilate physical states (this is just the statement that a physical state is gauge invariant). In conclusion we see that the area-entropy law (1. 4 Quantum gravity in de Sitter space So far we have discussed well established and understood results about classical de Sitter space and quantum ﬁeld theory in a ﬁxed de Sitter background. while a gauge transformation which goes to a constant at spatial inﬁnity will act nontrivially on the states. Gauge Theory and Strings Schwarzschild-de Sitter SSdS = π √ AH = 1 − 8GE. The results of this section are largely contained in [27] except for the derivation of the asymptotic boundary conditions for dS 3 . 11. (4. Parallel results pertain in arbitrary dimension.446 Unity from Duality: Gravity. by Noether’s theorem. about which little is established or understood.1 Asymptotic symmetries Consider a simple U (1) gauge theory in ﬂat Minkowski space. 4.39) The integration constant has been chosen so that the entropy vanishes for 1 the maximal energy E = 8G at which value the deﬁcit angle is 2π and the space has closed up. which were assumed/guessed without derivation in [27]. We will argue that quantum gravity in dS 3 can be described by a two dimensional conformal ﬁeld theory. 48]). Now we turn to the more challenging problem of quantum gravity in de Sitter space. 4G 2G (3. .1) where Oφ is an operator in the CFT associated to the ﬁeld φ. but the three dimensional case is the richest because of the inﬁnite dimensional nature of the I ± conformal group. 14. 47.

Since this group acts on I ± . Then the energy . and in principle there could be more than one choice.2) Here “allowed” means that the transformation is consistent with the boundary conditions that we have speciﬁed for the ﬁelds in the theory. which is deﬁned as the set of allowed symmetry transformations modulo the set of trivial symmetry transformations. one might try to demand that the metric fall oﬀ spatially as r12 . Therefore we need to consider diﬀeomorphisms in dS 3 which preserve the boundary conditions on the metric at I ± but do not fall oﬀ so fast that they act trivially on physical states.1) is simply an expression of diﬀeomorphism invariance of the theory. ASG = Allowed Symmetry Transformations · Trivial Symmetry Transformations (4. the theory will become trivial. Although we will not learn anything about the details of this theory. For example in 4d gravity.: De Sitter Space 447 It is useful therefore to consider the so-called asymptotic symmetry group (ASG). The states and correlators of the theory clearly must lie in representations of the ASG. However if the boundary conditions are too restrictive. but in some cases a knowledge of the ASG already places strong constraints on the theory. this means that correlators with points on I ± are those of a conformal ﬁeld theory.A. In this section we will see that the ASG of quantum gravity in dS 3 is the Euclidean conformal group in two dimensions. The result for de Sitter diﬀers only by a few signs. Strominger et al. 4. Of course one must know the details of the theory to know which representations of the ASG actually appear. On the 1 other hand one might try to demand that it fall oﬀ as √r .2 De Sitter boundary conditions and the conformal group Our ﬁrst task is to specify the boundary conditions appropriate for an asymptotically dS 3 spacetime. and “trivial” means that the generator of the transformation vanishes after we have implemented the constraints–for example asymptotically vanishing gauge transformations in the example of the previous paragraph. In quantum gravity the relevant gauge symmetry is diﬀeomorphism invariance. and in de Sitter space the only asymptotia are I ± . the fact that the conformal group in two dimensions is inﬁnite dimensional already strongly constrains the physics. and remains to be fully understood. In general speciﬁcation of the boundary conditions is part of the deﬁnition of the theory. However the physical interpretation in the dS 3 case is very diﬀerent from that of AdS 3 . This would allow only zero energy conﬁgurations and hence the theory would be trivial. and the correspondence (4. The analogous problem for three dimensional anti-de Sitter space was solved long ago by Brown & Henneaux [61].

but their results can be extended to the spacelike case. So the idea is to make the falloﬀ as weak as possible while still maintaining ﬁniteness of the generators.3) Here γ is the induced metric on the boundary I − and K is the trace of the extrinsic curvature Kµν = −∇(µ nν) = − 1 Ln γµν with nµ the outward2 pointing unit normal. are ﬁnite. = 3 Brown and York mainly consider a timelike boundary. Gauge Theory and Strings and other symmetry generators are in general divergent. The Brown-York stress tensor for dS 3 with = 1 is given by4 Tµν = 1 [Kµν − (K + 1)γµν ] . Hence we need to understand the surface integrals which generate the diffeomorphisms of dS 3 . 66]). . 2 = −1 + O(e2t ). z (4.4) For a perturbed metric gµν + hµν we obtain the Brown-York stress tensor Tzz = Tz z = ¯ 1 1 hzz − ∂z htz + ∂t hzz + O(h2 ). and hence all the symmetry generators. Requiring the stress tensor to be ﬁnite evidently leads to the boundary conditions gz z ¯ gtt gzz e−2t + O(1).5) Details of this calculation are given in Appendix A. and it is unlikely any sense can be made of the theory. (4. = O(1).448 Unity from Duality: Gravity. 4 We caution the reader that the generalization of (4. They showed that bulk diﬀeomorphisms are generated by appropriate moments of a certain stress tensor which lives on the boundary of the spacetime3 .3) to d > 3 or to theories with matter is not entirely straightforward [66]. (4. A convenient and elegant formalism for this purpose was developed by Brown & York [62. 4G 2 1 1 −2t 1 e htt − hzz + (∂z htz + ∂z ht¯ − ∂t hzz ) ¯ ¯ z ¯ 4G 4 2 + O(h2 ). 4G (4.3) vanishes identically for vacuum dS 3 in planar coordinates ds2 = −dt2 + e−2t dzd¯. We will deﬁne an asymptotically dS 3 spacetime to be one for which the associated stress tensor. 63] (and applied to AdS3 in [65.

6 Parallel derivations of the central charge for AdS were given in [64. The last boundary condition (4.7) where U = U (z) is holomorphic in z 5 . 8G (4. and we have not done so here.6) is violated very near the singularity). z) is an arbitrary function. In other words. 2G (4.: De Sitter Space gtz = O(1). in the sense described above.6) It is not hard to see that the most general diﬀeomorphism ζ which preserves the boundary conditions (4. The resolution of this apparent discrepancy comes from noting that if gtz → f on the boundary where f = f (z. The third term is the familiar linearization of the anomalous Schwarzian derivative term corresponding to a central charge c= 3l . if gtz = O(1). = O(1). 2 (4.7) acts on the Brown-York stress tensor as δζ Tzz = −U ∂Tzz − 2U Tzz − 1 U . In principle this should be carefully justiﬁed (as (4.10) 5 We allow isolated poles in z. . then in fact one can always choose a gauge in which gtz = O(e2t ). Exploiting this freedom one can impose the asymptotic boundary conditions gz z ¯ gtt gzz gtz e−2t + O(1). Therefore one can ¯ always set the component gtz of the metric to be O(e2t ) with a trivial diﬀeomorphism.8) The ﬁrst two terms are those appropriate for an operator of scaling dimension two. A parallel issue arises in AdS3 /CFT2 .7) do not contribute in (4. We conclude that the asymptotic symmetry group of dS 3 as generated by (4.9) where we have restored the power of 6 .6) may be written as 1 ζ = U ∂z + U ∂t + O(e2t ) + complex conjugate.6) diﬀers from the condition gtz = O(e2t ) assumed in [27] and obtained by analytically continuing the AdS 3 boundary conditions of Brown & Henneaux [61] from anti-de Sitter to de Sitter space. = O(e2t ).8).7) is the conformal group of the Euclidean plane. Therefore they are trivial diﬀeomorphisms. Strominger et al.A. 449 (4. Note that the O(e2t ) terms in (4. A diﬀeomorphism of the form (4. = (4. 2 = −1 + O(e2t ). 65]. then applying ¯ the diﬀeomorphism ζ = e2t f ∂z gives δζ gtz = O(e2t ).

and the dS 3 metric is therefore invariant.7) does. ¯ dz d¯ z (4.13) with φ chosen so that gzz → gz z ¯ ¯ under the combined transformation.4) makes it clear that this can be compensated by a shift in t. which acts like a Weyl transformation. The isometry group is the SL(2. which acts like an ordinary diﬀeomorphism of the complex plane. which from the point of view of the z-plane is a Weyl transformation. since in higher dimensional de Sitter space the ASG is the same as the isometry group SO(d. γ are complex constants. This is particular to the three dimensional case. So 2 a diﬀeomorphism in dS 3 splits into a tangential piece. (a) Find an example of string theory on de Sitter space. In two dimensions a diﬀeomorphism acts as gz z → ¯ and a Weyl transformation acts as gzz → e2φ gzz . (4. ¯ ¯ (4.12) followed by a Weyl transformation (4. A three dimensional diﬀeomorphism is thereby equivalent to a two dimensional conformal transformation.13) z dz d¯ gz z . Gauge Theory and Strings as given in [27]. Where do conformal transformations come from? Recall that a conformal transformation in two dimensions is a combination of an ordinary diﬀeomorphism and a Weyl transformation. . These transformations generate the SL(2.12) A conformal transformation is just an ordinary diﬀeomorphism (4. C) subgroup of the asymptotic symmetry group. C) global isometries of dS 3 . we conclude that the asymptotic symmetry group of gravity in dS 3 is the conformal group of the complex plane. This accounts for the second term 1 U ∂t in ζ. and a normal piece. A special case of (4.450 Unity from Duality: Gravity. Now if we look at what the diﬀeomorphism ζ deﬁned in (4. (b) Find the dual conformal ﬁeld theory. β. we see that the ﬁrst term U ∂z generates a holomorphic diﬀeomorphism of the plane. In this case U vanishes. Now the form of the metric (4. the ASG is inﬁnite dimensional. Since U (z) was arbitrary. a fact which highly constrains quantum gravity on dS 3 . We conclude these lectures with a last Exercise 4. In particular.7) is the choice U = α + βz + γz 2 . 1).11) where α.

7) and (A. 0 2 (A. 0. N z = N 1 −1 − htt . We can put (A.3) gives the desired result (4.A. 2e2t ht¯. z ¯ N z = 2e2t htz .4) The outward pointing unit normal vector to the boundary is nµ = 1 ¯ −1.: De Sitter Space A Calculation of the Brown-York stress tensor 451 We wish to calculate the Brown-York stress tensor (4. Strominger et al. N z . (A. ¯ z ¯ 2 2 2 (A.8) Plugging (A.1) and we will always drop terms of order O(h2 ). ¯¯ ¯¯ (A.5) Upon lowering the indices. ¯ ¯ z ¯ (A. = = 1 −∂z htz + ∂t hzz .2) (A.8) into (4. where the lapse and shift functions are given by 1 N = 1 − htt .6) and we use the formula Kµν = − 1 (∇µ nµ + ∇ν nµ ) to obtain 2 Kzz Kz z ¯ The trace is K = g µν Kµν = γ ij Kij = −2 − htt + 4e2t hzz − 2e2t (∂z htz + ∂z ht¯ − ∂t hzz ) . ¯ 2 γzz = hzz . 2e2t htz z 2 · (A. z.5). we have nµ = 1 1 − htt .3) and the induced metric on the boundary I − is γzz = hzz . z (A. 2 1 1 1 − e−2t (1 + htt ) − (∂z htz + ∂z ht¯ − ∂t hzz ) .1) into the form ds2 = −N 2 dt2 + γij (dxi + N i dt)(dxj + N j dt). We write the metric in planar coordinates (t.3) for a metric which is a small perturbation of dS 3 . 2 N z = 2e2t ht¯. z) as ¯ ds2 = gµν dxµ dxν = −dt2 + e−2t dzd¯ + hµν dxµ dxν .7) . γzz = ¯ 1 −2t e + hz z .

Hawking. Math. The asymptotic symmetry of de Sitter spacetime [hep-th/0107091]. Lett. Tsamis and R. Gao. Witten. B 453 (1999) 222 [hep-th/9901147].D.E. Y. Lin and Y. Symmetries.T. P. Klemm. D. A. Horava and D. Rev. Commun. Rev. Maldacena and A. Annals Phys. Rev. D. W. Phys. F. 38 (1998) 1113 [hep-th/9711200]. M. R. D.452 Unity from Duality: Gravity. R. Epstein and U. quantum mechanics and the cosmological constant [hep-th/0107031]. J. Vafa. Int. D 59 (1999) 047503 [hep-th/9810169]. Cosmological breaking of supersymmetry or little Lambda goes back to the future. and de Sitter gravity [hep-th/0107067]. 98 (1985) 105. Quantum gravity in de Sitter space. K. Khuri. Strominger and C. Phys. Strominger. B 440 (1998) 275 [hep-th/9806119]. A. Lambert. Strominger. C. Brotz and M. Theor. Minic. . Rev. JHEP 0011 (2000) 038 [hep-th/0010252]. Theor. Phys. 2 (1998) 231. Bousso. matrices. Phys. Odintsov.I. 43 (1975) 199. Witten. B 379 (1996) 99 [hep-th/9601029]. A. M. Chamblin and N. A. Lett.M. Bros.W. Phys.C.M. T. Hull. Some aspects of the de Sitter/CFT correspondence [hep-th/0106247]. V. Strominger. JHEP 9802 (1998) 014 [gr-qc/9801096]. Discreteness in deSitter space and quantization of Kaehler manifolds [hep-th/0101176]. Deser and A. T. Lett. II [hep-th/0007146]. Park. Banks and W. Commun. Phys.W. S. Volovich. B 575 (2000) 231 [hep-th/9911082]. E. D 7 (1973) 2333. Adv. Phys. Moschella. Math. J. H. Waldron. M-theory observables for cosmological space-times [hep-th/0102077].F. J. E. Banados. Lett. Wu. Phys. M. Phys. J. P. G. Quantum gravity in de Sitter space [hep-th/0106109]. Hawking. (A)dS backgrounds from asymmetric orientifolds [hep-th/0106209]. Phys. Phys. B 607 (2001) 577 [hep-th/0103198]. The dS/CFT correspondence [hep-th/0106113]. Ortiz. Woodard. Waldron. Li. D 15 (1977) 2738. JHEP 9906 (1999) 028 [hep-th/9906022]. Fischler. S. Phys. R. S. Math. Maldacena. N. C. Bousso.W. Phys. Gauge Theory and Strings References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] J. Banks. S. Nucl. Balasubramanian.res. D 59 (1999) 046002 [hep-th/9807216]. Nojiri and S.in/strings/Proceedings. Matrix model for de Sitter [hep-th/0106184].tifr. van Nieuwenhuizen and M. Pilch. Strings 2001 online proceedings http://theory. Deser and A. Zero-branes. Maldacena and A. Bousso. JHEP 9807 (1998) 021 [hep-th/9806146]. Conformal anomaly from dS/CFT correspondence [hep-th/0106191]. B 513 (2001) 137 [hep-th/0105181]. Hull and R. Kim. J. JHEP 0105 (2001) 001 [hep-th/0002145]. 253 (1997) 1 [hep-ph/9602316]. Hawking. Bekenstein. JHEP 0104 (2001) 035 [hep-th/0012052]. Phys.P. Silverstein. JHEP 0105 (2001) 043 [hep-th/0103171]. Sohnius. Nucl. Gibbons and S. T. S. E. R.

Halyo. [45] S.H.C. 44 (1975) 265. 114 (1978) 332. Nojiri and S. Math. Tagirov.W. Witten.. Phys.N. Phys. Perlmutter et al. ApJ 507 (1998) 46 [astro-ph/9805200]. D 30 (1984) 1733. dS/CFT correspondence and cosmic no-hair [hep-th/0109057].A. Rev. C. Shiromizu. Chernikov and E. Dolan.J. Annals Phys.D. Gravitational energy. Nappi.D. D 32 (1985) 3136.P. Math.R.W. 76 (1973) 561. Ottewill. Riess et al. [48] P.A. [59] B. Commun. [66] S. [54] H. Skenderis. Rev. Allen and A. Ida and T. Phys. York. Phys. [37] I. Poincare Phys. Kraus. [40] L. Rev.K. Deser and R.O. [60] B. Int. Davies. Phys. [43] B. Hoegh-Krohn and C. Tagirov.R. Phys. Mod. Conformal operators for partially massless states [hep-th/0109096]. Rev. Birrell and P.D. J. Deser. Solodukhin and K. . Comment on Nongaussian isocurvature perturbations from inﬂation [astro-ph/9705200].C. 217 (2001) 595 [hep-th/0002230]. Math. [44] A. P. Hull.A. Math.W. [53] R. A 15S1 (2000) 715 [eConfC 990809 (2000) 715]. Quantum Fields in Curved Space (Cambridge University Press. Ford. Folacci. Mottola. [41] R. inﬂationary brane-world creation and dS/CFT correspondence [hep-th/0107134]. Phys. Mazur and E. [39] T. Brown. [51] N. B 195 (1982) 76. Nucl. Phys. Lau and J.D. Abbott and S. Mazur and E. Phys. Rumpf and H. [64] M. [63] J. Allen. Ann. Perlmutter. [52] E. D 47 (1993) 1407. Phys. Phys. Henneaux. [46] S. JHEP 9807 (1998) 023 [hep-th/9806087]. Kallosh. Quantization of the massless minimally coupled scalar ﬁeld and the dS/CFT correspondence [hep-th/0108119]. A 9 (1968) 109. 1982). Odintsov. Turok. Phys. Cambridge.M. D 35 (1987) 3771. R.. [55] L. Tolley and N. Rev. Commun. Solodukhin. [36] E. Schmidt et al. Torii. [56] A.: De Sitter Space 453 [35] S. Brown and J.G.O. Quantum cosmology. [58] E. N. Nappi and E. 153 (1984) 405. Balasubramanian and P. Brown and M. S. Urbantke. [42] C. Henningson and K. Phys. Antoniadis. [61] J. Horizon holography [hep-th/0107173]. [65] V. de Haro. Ann. [38] A. Ann. [47] I. Commun. D 64 (2001) 104022 [hep-th/0106151]. D. Rev.A. Jackiw. De Sitter entropy and strings [hep-th/0107169]. Mottola. D 31 (1985) 710. Phys. Sachs and S.R. N = 2 supersymmetry and de Sitter space [hep-th/0109168]. Commun. Mottola. York. [62] J. [50] N. D.F. Phys. Theor.H. ApJ 517 (1999) 565 [astro-ph/9812133]. Figari. Strominger et al. Action and Energy of the Gravitational Field [gr-qc/0010024]. D 31 (1985) 754. [57] L. De Sitter Space in Supergravity and M Theory [hep-th/0109213]. S. [49] S. Rev. Skenderis. Najmi and A. 104 (1986) 207. ApJ 116 (1998) 1009 [astro-ph/9805201].. 208 (1999) 413 [hep-th/9902121].

35 route de Chartres.LECTURE 7 STRING COMPACTIFICATION WITH N = 1 SUPERSYMMETRY M. DOUGLAS Rutgers and IHES. France .R. 91440 Bures-sur-Yvette.

Contents 1 Introduction and the question of vacuum selection 2 From ten dimensional geometry to four dimensional eﬀective ﬁeld theory 3 D-branes with stringy corrections 4 Quantum corrections 5 Towards the low energy theory of everything 457 461 466 470 472 .

R. it had said little about the main physical questions identiﬁed in previous work. it is not in obvious crisis. string theory today appears relatively stable. and the relevance of noncommutative ﬁeld theory. solutions of gauge theory using brane techniques. This changed at the end of 1995 with the Strominger-Vafa derivation of the Bekenstein-Hawking entropy formula from string theory. were in 1995 either curiosities known to few (such as Dirichlet branes) or had not been conceived of at all. In the six years since. Although string theory was by then widely accepted as the leading candidate theory of quantum gravity.STRING COMPACTIFICATION WITH N = 1 SUPERSYMMETRY M. Although compared to 1994–97. the AdS/CFT correspondence. and some new and radically diﬀerent pictures of space-time. following the paradigms of heterotic string compactiﬁcation on a Calabi-Yau manifold. Even more surprising discoveries were to follow. in what is often called the “second superstring revolution”. Springer-Verlag 2002 . and the more recent Dirichlet brane paradigm. This situation is very good from the point of view of the c EDP Sciences. there is still much to do. Douglas Abstract We give a very broad overview of problems and recent work in the study of string compactiﬁcation with N = 1 supersymmetry. such as M(atrix) theory. We now have some non-perturbative understanding of the theory. and the ﬁeld remains very active and lively. Many of the central topics of our 2001 school. 1 Introduction and the question of vacuum selection The last Les Houches summer school which treated the fundamental theories of physics: gauge theory. the subject has changed dramatically. The new ideas have still not been completely understood and absorbed. was held in 1995 [29]. By 1997 the main ideas of this second revolution had emerged. gravity and the string and M theories which contain both of these. Much of the attention has shifted towards possible application to the real world. F theory.

It does seem to me however that too much importance . and from this point of view one might ﬁnd the school a bit too cut and dried. On the other hand. where opinions diﬀer is on how. a particular theory and internal space. which states that if choices enter in deriving predictions from string/M theory. but attempts to make more speciﬁc statements foundered on the problem that these depend on the choice of manifold. Just as importantly. At present there are not even good speculations as to how many four-dimensional compactiﬁcations of string/M theory there might be. At one extreme is the idea is that there is some a priori “Vacuum Selection Principle” waiting to be discovered. So let us try to look ahead a bit to the “third superstring revolution”. It seems to me that the third superstring revolution will have to address the problem of vacuum selection. after the primary test of consistency. Gauge Theory and Strings organizers of a school. But there need be no single principle which in itself governs the problem. and all of the couplings. one should not rule out possibilities a priori. one simply needs to sort through all possibilities and apply a series of tests to each. From this point of view. The original scenario of E8 × E8 heterotic string theory compactiﬁed on a six dimensional Calabi-Yau manifold led quite directly to theories resembling the standard model. but instead check whether observations can be ﬁt by any of the possible choices.458 Unity from Duality: Gravity. students (and everyone else) also come looking for vision. Far from helping. the second superstring revolution only seems to make this problem worse. which will uniquely ﬁx four macroscopic dimensions. as the outstanding barrier to contact with observation. each check for a possible conﬂict with observation could be regarded as a vacuum selection principle. the heterotic bound on the rank of the gauge group is easy to exceed. Many new ways were found to get gauge symmetry. as it allowed us to make a well motivated choice of topics. Perhaps the most basic question from the ﬁrst revolution left untouched by the developments of the second. and there are thousands if not millions (if indeed the number is ﬁnite!) of such manifolds. the speakers could feel that their time was well spent. which clearly enlarge the number of models. in introducing material of lasting importance (we think). and we have strong coupling duals for enough examples which satisfy these conditions to believe they are suﬃcient. Of course these are two extreme points of view and the truth is likely to lie in between. At the other extreme is the null hypothesis. I think many would agree with this. Early hopes that many compactiﬁcations would be inconsistent at the nonperturbative level now seem ill-founded. we know weak coupling nonperturbative consistency conditions for backgrounds. For example. is the problem of vacuum selection.

See [7. 11. and it seems a priori implausible that computing it in this way would lead to an acceptably small result. since the vacuum energy depends on all of these. The only generally accepted selection principle is the following: candidate vacua must be local minima of the vacuum energy. and it might be that a mechanism ﬁxing this will only work in one or a few backgrounds. Another likely arena for a selection principle to operate is cosmology.M. Having the cosmological constant as a boundary condition facilitates ﬁxing it by quantum or even anthropic considerations. Such mechanisms seem likely to work in a broad class of compactiﬁcations but not all. Clearly this is an outstanding problem. which is not at all necessary for string theory to make contact with reality. Even without these further considerations. It is certainly clear that one has the “antiselection” principle that putative vacua can be unstable to tunnelling to other nearby vacua.R. Douglas: N = 1 Supersymmetry 459 is given to the ﬁrst point of view. A diﬀerent realization of this idea emerges in the context of higher dimensional theories with gravity conﬁned to a four-dimensional brane [37]. while appearing as a parameter in the four dimensional Lagrangian. There are various ways that the hypothesis of a selection principle might be realized in string theory. This is an extremely powerful principle which one expects will stabilize all of the continuous parameters in the standard model Lagrangian. any instabilities or time dependence